A few weeks later, I googled “ross and pernilla wedding” to try to find their wedding website.

Google’s AI Overview responded confidently: Pernilla and Ross recently celebrated their romantic wedding ceremony and reception at the Royal Botanic Gardens, Kew in London... complete with a link to Kew and a citation trail leading back to... a guest’s Instagram posts.

Let that sink in. Two people with no website, no press coverage, no SEO strategy, and no commercial footprint whatsoever now have a synthesized, authoritative-sounding search result. Built from a handful of social posts by their wedding guests.

The discoverability threshold just got reset. And most brands haven’t noticed.

Subscribe now

The old threshold was brutal

For twenty years, being “discoverable” on the open web required scale. Domain authority. Backlink profiles. Years of compounding content. The SEO game was structurally winner-take-all: the top 10 results captured roughly two-thirds of all clicks, and breaking into them meant outspending or outlasting incumbents.

If you were small, new, or niche, the honest advice was: don’t bother competing on discovery. Buy ads instead.

A new threshold

AI-generated answers work differently. They construct answers, pulling from whatever signals are credible-enough on a topic. And the citation data reveals how radically that changes the game.

Evertune’s analysis of 200 million prompts found that no single domain dominates AI citations: even the most-cited domain rarely exceeds 5% of citations on any platform. The remaining 95% might scatter across thousands of domains. AI search has a long-tail distribution instead of a winner-take-all market.

Meanwhile, social content is becoming first-class source material. Tinuiti’s Q1 2026 data shows social media’s share of AI citations climbing steadily, reaching 13% for Google AI Overviews, and research from BrightEdge shows AI Overviews routinely citing Instagram, TikTok, YouTube and Reddit alongside traditional sources.

Ross and Pernilla didn’t make any conscious effort to out-SEO anyone. They simply became the key corroborated answer to a question. A couple of guest posts cleared the bar.

Why this is rocket fuel for the strategic few

Here’s the asymmetry worth exploiting: SEO rewarded accumulated authority; AI answers reward available, specific, corroborated signal. That collapses the time-to-discoverability from years to weeks.

For a startup, a local business, a niche B2B product, or a personal brand, this means:

Your category’s long-tail questions are winnable now. Nobody owns “best [your niche] for [specific situation]” in an AI answer yet. The specific beats the established.

Social is a vital source layer. Instagram posts, YouTube videos, LinkedIn articles, Reddit threads: AI systems treat all of it as citable evidence. A genuine presence in two or three of these can constitute the entire evidence base an AI draws on you.

Third-party voices matter. Notice that the wedding overview wasn’t built from anything Ross and Pernilla published, it was built from what other people said about them. Customer posts, community mentions, and guest commentary are now more structurally discoverable than ever.

Speed matters more than authority. AI systems synthesize fresh signal fast. The May 2026 wedding was in an AI Overview within two weeks. No domain-authority moat protects incumbents from a competitor who simply generates better source material, faster.

The takeaway

We used to ask: “How do we rank?” The new question is: “When an AI assembles the answer about our category, what evidence exists for it to find… and who created it?”

Two wedding guests with iPhones just demonstrated that the answer can be: almost nothing, and almost anyone.

If a handful of Instagram posts can make a private wedding discoverable to the world’s largest search engine, imagine what a deliberate strategy could do for your brand.

Leave a comment

For years, digital marketers have been trapped in a toxic relationship. The Meta-Google duopoly has held a chokehold on our budgets, forcing brands to endure skyrocketing CPMs, black-box algorithm updates, and dwindling returns. We’ve been begging for a true third competitor to break the compression.

Well, the alternative is finally knocking on the door.

Over the past 30 days, invitations to OpenAI’s self-serve ads beta have started dropping, giving select advertisers the ability to run ads directly inside ChatGPT. I was lucky enough to get early access and have been quietly playing around with the platform across a few of our brands for the last few weeks.

It is a massive milestone for high-intent marketing—but like any alpha-stage platform, it pairs groundbreaking potential with some seriously frustrating day-one operational friction. Here is exactly what it looks like inside the dashboard, the bugs you need to prepare for, and why this completely rewrites the playbook for Generative Engine Optimization (GEO).

Subscribe now

What OpenAI Got Right: The “Google Lite” Blueprint

OpenAI avoided a mistake that a lot of emerging ad networks make: they didn’t try to reinvent the wheel or overcomplicate the plumbing. They leaned into familiarity.

1. A Spartan, High-Intent Interface

When you first log into the dashboard, the experience is incredibly minimalist. It’s bare-bones and Spartan, which is a massive breath of fresh air. Usually, new platforms overwhelm you with complex navigation just to find simple things like your asset library or format settings.

Instead, OpenAI chose to mimic Google Search’s philosophy rather than Meta’s asset-heavy, visual dashboard. This is a brilliant move. ChatGPT isn’t a social network where you interrupt passive browsing with highly visual creatives; it’s a search destination. People go there to ask questions and solve problems. A clean, search-centric UI is exactly what media buyers need to capture that pure user intent.

2. Zero-Friction Tracking

Point number two: tracking data is dead simple. OpenAI requires you to drop a standard tracking pixel on your site to measure conversions and optimize campaigns. The script they gave us was a standard, familiar JavaScript snippet. Our engineers had it up and running immediately without any weird workarounds. By keeping the baseline tech identical to what the industry is used to, they’ve made it incredibly easy for brands to adopt it quickly.

3. Conversational “Context Hints” Instead of 500 Keywords

The campaign setup follows a traditional Campaign > Ad Group > Ad hierarchy, but it completely ditches the tedious nightmare of managing massive keyword lists or stressing over broad vs. phrase match syntax.

Instead, OpenAI implicitly signals that their LLM will figure out the matching logic for you. You just feed the system conversational prompts based on how real humans talk. You give it leading prompts like: Are they looking for the best running shorts for hot weather? Are they researching hair restoration treatments in New York?

You provide the natural language intent, and ChatGPT’s core engine handles the heavy lifting of mapping it to user queries.

The Early-Adopter Tax: The Friction and Frustrations

Now for the bad news. While the high-intent targeting is incredible, the operational plumbing is still very clunky.

1. The Agency Architecture Nightmare

Right now, when you apply for the beta, you define your website URL. The catch? Your ad account is permanently locked to only advertise on that exact domain. If your site is xyz.com, you cannot run ads for any other domain from that account.

If you are an agency or a media buyer managing multiple brands, this is a massive headache. You can’t centralize your workflows. You have to spin up entirely separate OpenAI beta accounts for every single client and wait for individual access approval for each one.

2. The Scraper Black Box (”Cannot Read Page”)

The most bizarre technical challenge we hit involved OpenAI’s web crawler. We launched an ad group and pointed the destination to a high-converting product page—a page that safely processes a couple hundred transactions a day.

Yet, the Ads Manager repeatedly blocked it with an error: “Cannot read page.”

Our engineers tore the page code apart and found absolutely nothing wrong. After a few days of banging our heads against a wall, we gave up and routed the traffic to our root homepage instead. For the first two days, the platform threw the exact same “cannot read page” error on our homepage, too. Then, suddenly, the crawler caught up, read the page, and the campaigns started delivering. Because ChatGPT ads rely heavily on the AI scraping your landing page to ensure contextual alignment with the user’s chat, a bug in their crawler can completely stall your launch.

3. Bare-Minimum Reporting

Don’t expect robust analytics or attribution suites out of the gate. Right now, the reporting dashboard is as Spartan as the UI. You are getting the absolute bare minimum data:

• Impressions

• Clicks

• Average CPC

• Daily Budget

Over time, they will undoubtedly build out a robust reporting suite, but right now, you are flying on pure aggregate metrics.

Why Paid Ads Make GEO More Critical Than Ever

The arrival of a paid ad ecosystem on LLMs doesn’t mean SEO or GEO is dead. In fact, it makes your organic optimization on these models hyper-critical.

To understand how paid and organic will live together on ChatGPT, I use a simple analogy:

Imagine you want to go see a movie in the theater. You look it up on Google, and right at the top, the search engine shows you that the film has a staggering 9.9 out of 10 rating on Rotten Tomatoes. Right next to that glowing, authoritative review is a prominent button that says “Buy Tickets Now.”

Because you just saw an objective validation of how great the movie is, you are infinitely more likely to click that ticket button and buy right then and there. If the review had been terrible, you’d ignore the button completely.

The Rotten Tomatoes score is the organic result; the ticket button is the paid ad.

This is exactly how paid ads and GEO will interact on LLMs. When a user asks ChatGPT, “What are the best running shorts for marathon training?” Your goal is to have your GEO and Answer Engine Optimization (AEO) dialed in so perfectly that the AI naturally lists your brand as an authoritative recommendation in its text response.

When your paid ad unit appears directly below that glowing organic recommendation, the conversion loop closes instantly. The LLM’s response provides the ultimate social proof, and your ad provides the immediate, frictionless path to purchase.

There is never a boring day in the world of search. As brands prepare to shift serious budgets into LLM advertising, the real winners won’t just be the ones who know how to buy the ads—it will be the ones who ensure their organic GEO work is done correctly so that the AI validates them first.

Leave a comment

I was filling out a signup form last week when a small detail stopped me cold.

The “How did you hear about us?” dropdown — arguably the most boring field in all of marketing — had a new option sitting at the very top of the list:

“AI tool (ChatGPT, etc.)”

Above YouTube. Above social media. Above Google.

Form fields are lagging indicators. Nobody adds a dropdown option for a channel that might matter someday. By the time something earns a permanent slot in your attribution survey, the behaviour behind it is already mainstream. Which means this unglamorous little checkbox is telling us something important: LLMs have quietly become a customer acquisition channel.

The ultimate dark traffic

“How did you hear about us?” exists because analytics can’t see everything. Word of mouth, podcasts, and recommendations have always been blind spots.

Historically, LLM referrals were that problem on steroids. Someone asks ChatGPT for “the best CRM for a 10-person agency,” gets a recommendation in a private thread, types your URL into a browser, and lands on your homepage. Your dashboard says: Direct. No referrer, no UTM, no click path. The most influential touchpoint in that buyer’s journey is completely invisible to you.

Until recently, this was almost entirely untrackable. Then, on May 7th this year, ChatGPT started surfacing clickable brand links directly inside its answers and the dam broke.

Similarweb measured a 157% week-over-week jump in ChatGPT referral traffic across tracked sites, with homepage referrals up more than 350%.

Roughly 60% of those visits now land on brand homepages, not deep content pages.

Small volume, outsized value

Let’s be honest about scale: for most websites, AI referrals are still a low single-digit percentage of traffic. If you’re judging this channel by volume, you’ll dismiss it.

You’d be wrong to. Similarweb’s clickstream data shows ChatGPT referrals converting at 7.1%, second only to paid search, and ahead of organic, social, email, and display. B2B agencies report AI-sourced visitors spending dramatically longer on site and converting meaningfully better than average. These buyers arrive with intent, because the “consideration phase” happened inside a chat window.

With ChatGPT at roughly 900 million weekly active users, even a thin slice of that behaviour shifting toward commercial questions is a serious acquisition force.

What this means for your stack

Three practical takeaways:

1. Add the dropdown option… today. It costs nothing and it’s currently the cheapest AI-attribution tool that exists. Self-reported attribution was treated as soft data for years; it’s now one of the few honest windows into AI-driven discovery. (Pro tip from the attribution nerds: don’t put it first in the list, people lazily click the top option, and you’ll pollute your own signal.)

2. Treat self-reported attribution as a first-class data source again. Pipe it into your CRM. Compare close rates and deal velocity for AI-tagged leads against other channels. Marketers who do this are consistently finding their dashboards under-report AI’s influence.

3. Start writing to be recommended, not just ranked. SEO is dead and a new discipline is forming around this call it GEO, AEO, or AI visibility (the naming wars are ongoing). The core shift: SEO optimised for a results page; this optimises for being the answer a model reaches for. Clear positioning, comparison content, authoritative third-party mentions, and structured product information all matter more than ever.

The checkbox is the canary

We spent two decades building an attribution industry on the assumption that customer journeys leave digital footprints. LLMs broke that assumption and the market’s first response wasn’t a new analytics platform. It was a humble dropdown option borrowed from 1995.

The companies adding that checkbox are admitting their dashboards have a blind spot big enough to deserve its own line item and getting ahead of the channel it represents.

Is “AI tool” in your dropdown yet?

Leave a comment

Remember all the hype over the past year around AI commerce?

The narrative was everywhere. Soon you’d ask ChatGPT what running shoes to buy, it would recommend the perfect pair, complete the checkout, and commerce as we knew it would fundamentally change. Entire conferences, startups, and venture decks were built around the idea that LLMs would become the next great commerce interface.

And honestly, the thesis made perfect sense.

Search was moving from links to answers. If AI becomes the interface layer between consumers and the internet, naturally it should also become the interface layer between consumers and purchasing.

So what happened? Why hasn’t commerce inside LLMs exploded yet?

The simple answer is that the frontier AI companies discovered a much faster path to monetization than commerce.

As OpenAI, Anthropic, Google, and others scaled usage, they also started seeing the terrifying reality of compute costs. Training these models was expensive. Running them at scale was even more expensive. The original assumption was that eventually advertising, recommendations, affiliate revenue, and transactional commerce would offset those costs.

But then something unexpected happened. Developers started willingly paying absurd amounts of money for productivity.

Claude Code becoming a standard workflow for developers changed the economics almost overnight. Suddenly companies were seeing engineers become dramatically more productive using these systems. Not marginally better. In some cases 2x to 5x more productive. Features shipped faster. Prototypes got built in hours instead of weeks. Smaller teams suddenly operated like much larger organizations.

And once companies saw that leverage, the willingness to pay became enormous.

That completely changed the roadmap for the LLM companies. Why spend years solving the incredibly difficult problems underneath AI commerce when enterprises were immediately willing to pay massive recurring revenue for AI productivity today?

Commerce suddenly looked hard. Enterprise looked immediate. That shift is massively underappreciated.

AI commerce requires solving trust, payments, logistics, recommendations, returns, fraud, attribution, merchant integrations, and autonomous agents that can reliably execute transactions without breaking. The infrastructure problem underneath commerce is enormous.

Enterprise AI, on the other hand, mostly required making expensive employees dramatically more productive. One creates immediate ROI. The other requires changing consumer behavior at scale. So the labs followed the money.

Instead of prioritizing conversational shopping and embedded checkout, they prioritized coding agents, APIs, reasoning, enterprise copilots, workflow automation, security, and developer tooling.

Then another thing happened. Companies started realizing they didn’t necessarily want all of their proprietary data and workflows sitting inside centralized AI systems forever. Developers started demanding more local control over compute, workflows, and inference. Projects like OpenClaw and the explosion of local AI tooling accelerated the idea that AI wasn’t just becoming software — it was becoming infrastructure.

That created an entirely new market around local inference, private models, AI workstations, enterprise-controlled compute, and self-hosted AI environments.

Again, commerce got pushed further down the roadmap.

So was AI commerce fake? No. Not even close. I actually think AI-driven commerce eventually becomes one of the biggest platform shifts in internet history. But the timing assumptions were wrong because everyone underestimated how quickly enterprise AI spending would materialize.

The AI labs thought they might become the next Google. Instead, they accidentally discovered they could become the next Microsoft, Oracle, AWS, and Accenture first.

Commerce is still coming. We are just in the infrastructure monetization phase of AI before the transactional phase fully arrives.

But if you run a commerce company, you should absolutely be preparing now. Because when AI agents do become trusted purchasing layers, distribution will change very quickly. If you run an e-commerce brand, here are the 5 things to do as soon as possible.

1. If you’re on Shopify, enable agentic commerce capabilities now.

Shopify is clearly positioning itself for AI-native shopping and autonomous checkout flows. Most merchants still have not enabled the infrastructure that will allow AI agents to interact with their stores intelligently. Early movers here will have an advantage.

2. Follow Google’s emerging universal commerce protocols for AI.

Structured product data is going to matter enormously in an AI-commerce world. LLMs cannot reliably recommend products from messy catalogs and inconsistent metadata. Brands that adopt machine-readable commerce standards early will be significantly easier for AI systems to understand and transact against.

3. Optimize your website for LLM readability, not just Google SEO.

Traditional SEO optimized for ranking pages. AI optimization is about helping reasoning engines understand your product, trustworthiness, pricing, policies, compatibility, fulfillment quality, and customer experience. The brands that communicate clearly to machines will outperform companies still relying on old-school keyword tactics.

4. Optimize for frictionless checkout.

If AI agents eventually complete purchases autonomously, friction becomes conversion death. Reduce unnecessary checkout steps, remove account friction, optimize mobile performance, and enable one-click payment systems wherever possible. The easier your infrastructure is for both humans and machines, the better positioned you’ll be.

5. Build review volume and customer sentiment aggressively.

AI systems will heavily weight trust signals including reviews, sentiment, reliability, satisfaction, and return rates. In the next generation of commerce, reputation becomes distribution. Brands should treat customer feedback and post-purchase experience as core infrastructure, not marketing support.

The next era of commerce may not be won by the companies with the biggest advertising budgets. It may be won by the companies most understandable, trustworthy, and executable by machines.

Leave a comment

Today we’re announcing a brand new addition to our Substack publication: the SEO is Dead subscriber chat!

This is a conversation space exclusively for subscribers, kind of like a group chat or live hangout. We’ll post questions and updates, and you can jump into the discussion.

In this forum, we’re your guide through the wave of disruption caused by AI and the shift from conventional search to AI-first platforms like ChatGPT — and what that really means for discovery, brand relevance, and customer acquisition.

Whether you’re confused by the world of AI or have questions you’ve been too embarrassed to ask out loud, consider this your safe space. We’ll cover everything from LLMs and SEO to AEO and the dark art of getting cited by AI.

We want to hear from you — so what would you like to know?

Join chat

How to get started

1. Get the Substack app by clicking this link or the button below. New chat threads won’t be sent sent via email, so turn on push notifications so you don’t miss conversation as it happens. You can also access chat on the web.

2. Open the app and tap the Chat icon. It looks like two bubbles in the bottom bar, and you’ll see a row for my chat inside.

3. That’s it! Jump into our thread to say hi, and if you have any issues, check out Substack’s FAQ

I’ve been through a lot of Google updates over the years.

I was around for Panda in 2011, when Google went after content farms and SEO spam. Then Penguin in 2012, which focused heavily on unnatural linking, keyword stuffing, and web spam. Back then, the internet was flooded with people trying to game the algorithm, and Google’s job was to clean it up.

Then came some of the biggest updates I personally saw impact the industry: Hummingbird and Pigeon in 2013 and 2014.

Hummingbird fundamentally changed Google’s core algorithm and pushed search toward semantic understanding instead of simple keyword matching. Pigeon dramatically improved local search and changed how businesses ranked geographically.

But here’s the important thing people miss when they look back at these updates:

Every major Google update historically followed a shift in user behavior. The early updates cleaned up desktop SEO spam. Then between 2015 and 2022, the focus shifted heavily toward mobile.

Mobilegeddon in 2015 basically told the world: if your website isn’t mobile-friendly, you’re not going to rank well anymore. Even on the desktop.

Then BERT in 2019 pushed Google even further into understanding context, intent, and natural language, especially on mobile experiences. Google spent almost a decade telling businesses: If you’re not optimized for mobile, you’re eventually going to lose visibility everywhere.

Now we’re entering the next era. And this recent core update made it very clear that if you’re not optimized for AI, then forget about SEO!

We’re Moving From a Mobile-First Internet to an AI-First Internet

The updates Google rolled out between February and April 2026 weren’t random volatility. This reminded me a lot of Panda and Penguin all over again. The focus now is AI spam instead of desktop spam.

More specifically: mass-produced, low-effort, AI-generated content. A lot of major publishers jumped onto the AI bandwagon over the last 36 months. They started scaling content production aggressively using AI-generated reviews, summaries, comparison pages, and thin informational articles. And during this update cycle, many of those sites got crushed.

We saw pages that historically ranked extremely well suddenly drop 20–30% or more, especially review sites and thin-content publishers where you could clearly tell the content lacked originality, firsthand expertise, or real value. Why? Because Google is prioritizing something very different now.

Google Is Rewarding Originality Again

What Google wants now is:

• proprietary insights

• original research

• firsthand experience

• demonstrated expertise

• real credibility

If you’re a researcher, operator, doctor, lawyer, investor, founder, engineer, or actual practitioner with firsthand experience and unique data, you now have a massive advantage over generic AI-generated content.

That’s the shift. For years, SEO rewarded scale. Now GEO rewards credibility. And the industries that got hit the hardest during this update were YMYL categories: Your Money or Your Life - Health, Finance, Legal.

Why? Because Google does not want AI-generated fluff giving people medical advice, financial advice, or legal guidance without real expertise behind it.

That’s where E-E-A-T becomes critical: Experience. Expertise. Authoritativeness. Trustworthiness.

And honestly, this is probably just the beginning. That means the future winners are not going to be the companies producing the most content. The winners are going to be the companies producing the most trusted content.

And trust now extends beyond your website.

Google and AI systems are increasingly evaluating:

• author credibility

• entity authority

• citations across the web

• consistency across platforms

• original data

• real-world expertise

• user trust signals

In other words: Google is no longer just ranking pages. It’s evaluating credibility systems.

What Businesses Should Do Going Forward

Honestly, the strategy is pretty simple.

Create original content. Show real data. Publish firsthand experience. Stop mass-producing low-effort AI pages hoping to rank. Use AI to amplify expertise, not fabricate it.

The latest core update confirmed exactly where Google is heading next. We’re moving from an SEO-first internet to a trust-first AI internet. The businesses that understand that early are going to dominate the next decade of search, AI Overviews and beyond.

Leave a comment

Subquadratic just built the LLM the industry said wasn’t possible. Here’s what it can actually do.

There is a limit inside every AI system you’ve ever used. You can’t see it, but it’s there: a hard architectural ceiling on how much a model can hold in its head at once. Feed it too much, and it starts forgetting. Push further, and the cost becomes prohibitive. Most of what passes for “AI” today is really artificial intelligence operating under severe memory constraints, propped up by a scaffolding of workarounds nobody particularly wants to talk about.

A Miami-based startup called Subquadratic thinks it’s broken through that ceiling.

On May 5th 2026, the company emerged from stealth with $29 million in seed funding and a single, very loud claim: it has built the first frontier large language model that doesn’t behave the way every other LLM does under the hood. Its model, SubQ, offers a 12 million token context window… at roughly one-fifth the cost of leading frontier models. If those numbers hold up at scale, this is the most significant architectural shift in large language models since the original transformer paper in 2017.

The Problem Nobody Was Solving

To understand what Subquadratic built, you need to understand what everyone else was living with.

Every mainstream AI model today, GPT, Claude, Gemini, is built on something called a transformer architecture. Transformers are extraordinarily capable, but they have a fundamental scaling problem: to understand language, they process every possible relationship between every word in the input. That relationship grows quadratically. Double the context length, and compute quadruples. At the scale of millions of tokens, the numbers become untenable.

The industry adapted by building around this limitation rather than solving it.

Retrieval-augmented generation (RAG) became the dominant workaround. Instead of feeding a model everything, you build a search layer that retrieves the most relevant fragments first, then passes those to the model. It’s effective. It’s also expensive, brittle, and fundamentally limiting. You’re not asking the AI to reason across all of your data. You’re asking it to guess which slice of your data matters, then reason across that.

As Subquadratic’s own team put it: developers spend more time and money on the workarounds than on the actual problem.

The Architectural Breakthrough

SubQ is built on what the company calls Subquadratic Sparse Attention (SSA). Rather than processing every possible relationship between tokens, SSA dynamically routes attention only to the relationships that actually matter, based on content. The result is a model that scales linearly rather than quadratically. At 12 million tokens, that difference reduces attention compute by almost 1,000 times.

In practical terms: SubQ runs 52 times faster than existing state-of-the-art attention mechanisms at one million tokens. On a standard long-context benchmark, it achieved comparable accuracy to Claude Opus at a cost of roughly $8 — versus approximately $2,600 for Opus on the same task. That’s a different cost category entirely.

The benchmarks are promising. On MRCR v2 — a demanding test of multi-round reasoning across long contexts — SubQ scored 86, outperforming Claude Opus (78), GPT-5.4 (37), and Gemini 3.1 Pro (26). Results were third-party validated, though independent external reproduction is still forthcoming.

What This Actually Unlocks

Here is where the architecture stops being a technical story and becomes a business one.

Legal and professional services: A firm running M&A due diligence today feeds documents into a retrieval system, retrieves fragments, and hopes the right clause surfaces. With a 12 million token context window, you feed in the entire data room — every contract, filing, email, disclosure — and ask a single question across all of it.

Software engineering: Today’s AI coding agents operate with a narrow slice of a codebase at a time. They make changes without full visibility into how those changes ripple through thousands of other files. SubQ’s Code product — designed to plug directly into Claude Code, Cursor, and Codex — lets an agent hold an entire repository in context simultaneously, alongside months of pull requests and test history.

Financial analysis: An analyst could load a company’s complete transaction history, every earnings call transcript from the past five years, and all regulatory filings simultaneously and ask nuanced questions across the entire evidence base.

The common thread across all of these: the workaround disappears. You stop engineering around the model’s memory limitations and start asking it the actual question.

The Bet

The team behind Subquadratic comes from Meta, Google, Oxford, Cambridge, and BYU. CEO Justin Dangel and CTO Alex Whedon (previously Head of Generative AI at Meta) founded the company on the thesis that the transformer’s quadratic bottleneck wasn’t a feature of intelligence, it was a constraint on it. Backers include Justin Mateen (Tinder co-founder) and Javier Villamizar, former partner at SoftBank Vision Fund.

The pre-seed valuation is reported at approximately $500 million, an extraordinary number for a company that launched weeks ago, and a signal of how seriously sophisticated investors are taking the architectural argument.

SubQ is currently in early access, with two products: a full-context API for developers and enterprise teams, and SubQ Code, the long-context layer for coding agents. The technical report is forthcoming.

The claims are bold. The benchmarks are promising. Independent verification is still catching up. But Subquadratic’s case is a simple one: they’ve built the first model designed from the ground up for long-context reasoning — and at a $500 million pre-seed valuation, investors are betting they’ve pulled it off.

Leave a comment

Brand migrations used to be an SEO problem. Now they’re an AI visibility problem too.

In the age of Generative Engine Optimization (GEO) and Answer Engine Optimization (AEO), changing your company name, domain, or URL structure can confuse not just Google — but also Large Language Models that power AI search and recommendations.

I have worked directly on brand name (or domain name) changes for some fairly large companies, and in the past, all we cared about was raising awareness of change and maintaining our SEO rank. However, now we have to ensure that our authority on the answer engines retain their old authority and develop new authority as well!

Here are the ten most important steps to manage a successful GEO-era migration:

1. Keep SEO Fundamentals Intact

Use 301 redirects, preserve URL structures, update sitemaps, and use Google Search Console’s Change of Address tools. Traditional SEO signals still matter.

2. Explicitly Connect the Old Brand to the New One

Don’t assume AI systems will infer the relationship. Clearly state:

• “Brand A is now Brand B”

• “Formerly known as…”

Repeat this across your website, press releases, and profiles.

3. Treat Your Website as the Primary Source of Truth

Your site remains the strongest authority signal for LLMs. Update:

• About pages

• Metadata

• Navigation /Footer references

• Structured data

Consistency matters more than creativity during migrations.

4. Update Structured Data Immediately

Use Organization schema, legal entity fields, and sameAs links to reinforce identity continuity across the web.

5. Fix Third-Party Profiles Fast

Update high-authority sources first:

• LinkedIn

• Crunchbase

• Industry directories

• Wikipedia (when applicable)

These platforms heavily influence AI understanding.

6. Use Digital PR to Reinforce the Change

Secure earned media coverage announcing:

• The old brand

• The new brand

• Why the transition happened

Repetition across trusted publications accelerates AI recognition.

7. Clean Up Legacy Mentions Across the Internet

Old Reddit threads, forum discussions, blogs, and review sites can continue reinforcing outdated identity signals. GEO migrations require active cleanup by either commenting on old content with corrections or reaching out to the blogs for updates.

8. Avoid Multiple Major Changes at Once

Changing your:

• Brand

• Domain

• Messaging

• Site architecture

…all simultaneously creates confusion for both search engines and AI systems. Phase changes carefully whenever possible.

9. Publish High-Authority Content Under the New Brand

Thought leadership accelerates authority transfer. Publish:

• Research

• White papers

• Expert commentary

• Industry insights

This helps LLMs associate expertise with the new entity.

10. Understand the Core Difference Between SEO and GEO

SEO migrations are about preserving rankings. GEO migrations are about preserving identity.

Google eventually reconciles redirects. AI systems require repeated, distributed validation across the web before they fully trust a new brand identity.

Share

A few days ago, I was at an investor conference focused almost entirely on consumer packaged goods: cereal, snacks, beverages, household staples. Traditional CPG.

And yet one of the dominant conversations wasn’t about pricing, supply chains, or retail expansion. It was AI!

More specifically, it was about how large language models are beginning to influence what consumers buy. One stat stood out above everything else: roughly 25% of consumers are already using AI tools to get recommendations for products they can purchase in stores. That number is likely headed much higher over the next few years.

That changes everything. For decades, brands fought to win at the shelf. Then they fought to win in search. Now they have to win in chat.

A Short History of Shopping

In the 1990s, shopping was physical. Consumers walked aisles and made decisions based on packaging, placement, and price.

In the 2000s, the internet began shaping decisions before shoppers entered the store. Search engines and retailer websites became influential.

In the 2010s, smartphones compressed the decision window even further. Consumers researched products in real time, often while standing in the aisle.

Now we’ve entered the next phase: conversational commerce.

Consumers increasingly skip search entirely and ask AI directly:

• “What’s the healthiest cereal?”

• “What’s a better alternative to this snack?”

• “Which pasta sauce has the cleanest ingredients?”

Instead of ten blue links, they get one synthesized answer.

That means the interface for discovery is no longer the shelf or the search engine.

It’s the answer layer.

1. Turn Your Website Into an Answer Engine

Most brand websites were designed for human browsing.

Now they also need to be readable by AI systems.

Brands should clearly explain:

• what a product is

• who it’s for

• why it’s different

• how it compares to alternatives

The clearer the explanation, the more likely it is to appear in AI-generated recommendations.

2. Structure Content Around Questions

Consumers interact with AI conversationally, not through keywords.

That means brands should create content that directly answers questions people actually ask:

• Is this healthy?

• Is it worth the price?

• What’s the best option for kids?

• How does this compare to competitors?

Brands that mirror natural language queries are more likely to surface in AI outputs.

3. Own the Comparison Narrative

If brands don’t explain how their products compare to competitors, AI models will do it for them.

That’s risky. The winning brands will proactively publish:

• comparisons

• rankings

• “best for” use cases

• nutritional breakdowns

• feature differentiation

In the LLM era, comparison content becomes strategic infrastructure. If your gun shy on having your competitors ranked on your own site, then you might as well give up now. You won’t win in the new world of Answer Engines!

4. Build Strong AI-Readable Entities

AI models don’t just recognize keywords. They build associations.The strongest brands consistently reinforce the attributes they want associated with their products:

• low sugar

• high protein

• premium ingredients

• affordable

• sustainable

• family-friendly

Over time, these associations shape how models describe and recommend products.

5. Optimize for Zero-Click Decisions

In many cases, the consumer will never visit your website.

The AI summary is the customer experience. That means your brand messaging needs to survive compression into a few sentences inside ChatGPT, Gemini, Claude, or other systems.

Brands now need to think not just about what humans read, but what machines summarize.

The Bottom Line

We are entering a world where a growing percentage of purchase decisions happen before a consumer reaches the aisle. That is a fundamental shift.

The brands that adapt earliest, by structuring information clearly, shaping AI-readable narratives, and optimizing for conversational discovery, will become the default recommendations inside the systems consumers increasingly trust.

For decades, brands competed for shelf space. Now they’re competing for answers.

Leave a comment

One of the reasons people got excited about platforms like OpenClaw is simple, it feels local.

What that means in practice is powerful. Instead of uploading your data and credentials to some distant AI service, you can keep everything on your own machine, your Gmail, Slack, Google Drive, Salesforce, even Apple services, all logged in locally. Then, you layer an AI agent on top to automate workflows, whether that’s scanning emails, triggering actions, or coordinating tasks across systems.

At first glance, this seems like the perfect balance, local control of sensitive data and credentials combined with powerful AI-driven automation. That’s exactly why tools like OpenClaw captured so much attention.

The Illusion of “Local” AI

But there’s a catch, and it’s an important one.

While platforms like OpenClaw operate locally in terms of access and orchestration, the intelligence itself is usually not local.

To actually perform meaningful reasoning, these systems still rely on external large language models (LLMs), such as those from OpenAI, Anthropic, or others. In practice, that means data is often sent to external servers, computation happens off-device, and results are returned back to your local environment.

So while your logins may stay local, your data flow does not fully stay local.

This creates a real concern: If your data still needs to leave your machine for intelligence, is it truly secure?

That tension, between local control and remote intelligence, is where the current generation of tools falls short.

A Look Back, When Software Was Truly Local

To understand where this could go, it helps to look back.

In the 80s, 90s, and early 2000s, software was fundamentally different. You installed tools like Microsoft Office directly onto your machine, and everything, from processing to storage, happened locally, with no reliance on the cloud or external services.

That model gave users full control over their data, eliminated the need for constant connectivity, and created a predictable, contained environment.

AI today has largely moved in the opposite direction, toward centralized, cloud-based intelligence, but that may not be where it ends.

The Shift Toward Fully Local AI

What’s emerging now is the possibility of true local AI, where both the data and the computation stay entirely on-device.

This shift is becoming more realistic thanks to smaller and more efficient models, advances in hardware like GPUs and Apple Silicon, and optimization techniques such as quantization and distillation. Together, these make it possible to run increasingly capable models locally without constant reliance on external infrastructure.

In this world, you wouldn’t need to send your data to any external LLM at all. Your AI assistant would run locally, process locally, and operate within your environment, meaning your data never leaves your machine.

Why This Matters, The Enterprise and Proprietary Data Use Case

The biggest impact of local AI won’t be consumer convenience, it will be enterprise trust.

Take a hospital, for example. Imagine decades of proprietary radiology reports, tens of thousands of highly sensitive records. Today, even with modern AI tools, there’s hesitation around whether any of that data is being sent externally, whether it could be used for training, and whether it fully complies with privacy regulations.

Even with safeguards, the concern remains.

Now imagine a different scenario, a fully local AI system that is pre-trained but deployed entirely on-premise, capable of analyzing those radiology reports without ever transmitting data outside the organization.

That changes everything.

The same dynamic applies across financial institutions, defense systems, legal firms, and any environment where proprietary or sensitive data is core to the business. In these cases, data privacy isn’t just a feature, it is the product.

Where the Real Competitive Moat Will Be

Today, many companies are building on top of LLMs, creating wrappers, agents, and workflow tools.

But LLM capabilities are rapidly converging. Open and commercial models are getting closer in performance, and switching between them is becoming easier. As a result, the long-term advantage is less likely to come from simply having the best model.

Instead, the real moat may come from the ability to run AI securely, locally, and effectively on proprietary data.

That’s a very different kind of advantage, and a much harder one to replicate.

The Next Wave, Local AI Appliances and Preconfigured Systems

Don’t be surprised if the next evolution looks something like this.

Instead of subscribing to a cloud service, you might purchase a pre-configured local AI system, perhaps costing $1,000 or more depending on its capabilities, designed for a specific use case such as healthcare analysis, enterprise workflows, or personal productivity.

You install it, or it arrives as a dedicated device, and it runs entirely locally. It’s optimized for your workflows, configured for your environment, and critically, it never sends your data back to any provider.

In many ways, this would mark a return to a familiar model, software you own, running on machines you control, now enhanced by AI.

Why Platforms Like OpenClaw Still Matter

This is where platforms like OpenClaw become especially important.

They represent an early step toward local orchestration, credential control, and workflow automation tied directly to real user environments. Even if they still depend on external models today, they serve as a bridge between today’s cloud-based AI and a future where everything runs locally.

As local models improve and hardware continues to advance, platforms like these could evolve into fully self-contained AI environments with no external dependencies.

Final Thought

We are in a transitional phase. Right now, AI is powerful, but mostly centralized. Tools like OpenClaw give us a glimpse of a more controlled, local-first future, but they haven’t fully crossed that boundary yet.

When true local AI arrives, where intelligence, data, and execution all live on your machine, it won’t just improve these platforms. It will redefine their value entirely.

And that’s when tools like OpenClaw, and everything built around them, become not just useful, but essential.

Leave a comment

How to Unlock Your Generative Engine Optimization (GEO) and Answer Engine Optimization (AEO).

At Literate AI, we’ve been in the AI trenches since before ChatGPT launched, getting nerdy on how semantic search and Natural Language Processing (NLP) make or break brands and reputations. Earlier, our founding team of Mitch Stoller, Matt Van Dusen and Gabe Gold launched and led Group SJR, the first pure-play content marketing agency.

So we have a strong view of what makes content work, how to get discovered in search, build a consumer/customer journey across your site, social, and more. But with the rise of the Chatbots everything’s changed. It’s no longer enough to think about what humans want. You need to understand the Bot’s Journey, too.

As a leader in AIO, GEO and AEO spaces—well at least the NYT, the Ankler, Sports Business Journal have quoted us and published our research—we wanted to share our principles for success. We’ve derived these rules from applying our geekiness to helping global enterprises, tech startups, governments and NGOs get discovered by Generative AI.

Here’s our Ten Laws of Generative Engine Optimization (GEO) and Answer Engine Optimization (AEO).

Law 1: AI is Your Brand’s Most Important Audience

Stop worrying about your human customers for a second. In a zero-click world, AI answers questions, not Search Engine Results Pages. It means you need to win over the Bots!

Law 2: Search is Snobby, AI is Hungry

The rise of AI Overviews and answers also means the reign of search snobbism is over. In the past, if you weren’t legacy media, you were sitting at the kid’s table. But AI is different. It’s hungry for deep, specific information to answer the next five follow-up questions a user might ask.

Law 3: Machines Don’t Read Like We Do

Maybe you think the robots are reading your blog posts. Wrong. They are doing math, and they don’t care about a hook that pulls you in. They break your words into ‘tokens’ and turn them into numerical IDs. It’s all vectors and probability.

Law 4: Write for Robots, Not Just People

Keep your storytelling for Substack. You need to be a direct answer machine. Stop flirting with the reader and start answering their follow-up question. Structure your content for the human eye, and you’re invisible to the algorithm. Structure it for the neural network, and you become the answer.

Law 5: Algorithms Are a Moving Target

Back in the Golden Age of SEO, Google rewarded great content with high search engine results page rank. Great blue links! Not anymore of course. But we’re in a similar Golden Moment for AI right now, but it won’t last. LLMs will eventually stop ingesting content for free and start charging you for discovery. If you aren’t already the ‘right answer’ when that shift happens, breaking into a chatbot response will be just as hard as landing page one on Google^.

Law 6: Know Thyself to Know Thy Queries

The machines are black boxes. They won’t tell you what people are asking. So, if you want the right answers, you have to find the right questions yourself.

Forget keyword volume. You need to map your own ‘Query List’—40 to 60 deep questions that define your brand and your category. This is how you discover what the AI doesn’t know about you. If you don’t shape the questions now, you’ll never control the answers later.

Law 7: Your Website is Your Most Valuable Digital Asset

Think your website is for customers? Think again. In the age of AI, your website is actually a training manual for the robots. Stop obsessing over ‘user journeys’ and start building ‘bot journeys. Use your owned digital property to feed the LLMs. Structure your sentences with positive sentiment and isolate the negatives. If you aren’t providing the data, the AI will just guess

Law 8: Map the Bot’s Journey

Marketing since 1898 has been about AIDA: Attention, Interest, Desire, and Action. But here’s the truth: Chatbots don’t have desires. They have algorithms. The consumer journey is dead. You need to map the Bot’s Journey. It’s not about emotional persuasion; it’s about predictive patterns and data sequences.

To win, you have to get RADD: Relevance, Authority, Discovery, and Depth. The bot is looking for HTML over PDFs and depth over catchy headlines. If you want to be the answer, you have to understand where the bot looks and in what order. If you aren’t RADD, you’re forgotten.

Law 9: Winner Won’t Take All

For 20 years, Google was the only game in town. But in the AI age, the ‘Winner Take All’ world is officially over. Google is cannibalizing its own search business just to keep up, while Perplexity and Claude are carving out massive territories in research and code.

You can no longer rely on a single silo. Even Grokipedia is rising to challenge Wikipedia’s dominance over AI training data.

Your brand needs to be everywhere the bots are. If you’re only optimized for one channel, you’re missing the rest of the world. Stop betting on one horse.

Law 10: Orchestrate, Don’t Dictate

The final law to rule them all: Orchestrate, don’t dictate. Marketing in the AI age isn’t about managing silos; it’s about leading a symphony.

Forget simple automation. Orchestration means using AI to sync every customer touchpoint in real-time—from SMS to LLM answers.

Automation is doing things right. Orchestration is doing the right thing at the exact moment your customer needs it.

McKinsey projects this market will hit $1 trillion by 2030. If you don’t start thinking holistically today, you’ll be left behind tomorrow.

Leave a comment

I’ve been spending a lot of time lately experimenting with different AI coding and agent-style tools, and as an engineer who actually uses this stuff day to day, I tend to care less about hype and more about what holds up after a few real workflows.

OpenClaw was one of those projects that caught my attention early because it showed something important: you can start to stitch together a more autonomous, local-first development experience. That said, after putting OpenClaw through its paces, I found myself reaching for a handful of other tools more consistently. Here are six practical replacements for OpenClaw available now, each with a different twist.

Perplexity Computer is the closest thing to a thinking layer in the cloud that directly takes on OpenClaw. It’s strong at breaking down problems, comparing strategies, and helping you discuss decisions before you even touch code. It ends up being more useful than expected because it reduces bad starts, which is something most AI Agent tools don’t address. One of the biggest gripes from OpenClaw users is bad starts. Users have to spend hours setting OpenClaw up and then even more time setting up the right work flow.

Tasklet.ai is more execution-focused. It’s not trying to be everything. It’s trying to help you actually get repetitive or structured tasks done without spinning up full systems. When I want to chain together small workflows or automate parts of a process that are too minor to justify building out but too frequent to ignore, it comes in handy. It sits nicely in that middle ground between scripting and full automation. Tasklet recently got $20m in funding from leading VCs; so a lot of smart people see a future for this company.

Base44 has an experimental edge, but it’s still productive. It’s the kind of tool you reach for when simple prompt-in, answer-out starts to feel limiting. It gives you more flexibility to shape interactions and build more complex flows without a ton of overhead. It’s not the most polished, but it’s definitely one of the more interesting OpenClaw alternatives if you like pushing boundaries.

Claude Code is one of the more dependable tools in my workflow. It handles larger context better than most, which matters when you’re working across multiple files or trying to maintain consistency. When I need actual reasoning and not just suggestions, especially for refactoring or understanding unfamiliar code, I use it. It’s not flashy, but it’s consistently useful. But one challenge with Claude Code vis-a-vis OpenClaw is that it’s designed primarily for engineers, making it hard to setup workflows .

Cursor AI is where things start to feel embedded into your actual development environment. Instead of stepping out to ask questions, it works alongside you while you code. It understands the context of your project as you go, which changes how you write and iterate. It’s less about one-off help and more about continuous assistance, which makes a noticeable difference over time. Again, a bit heavy on the coding side, but you can set up workflows that are more complex - think enterprise level workflows in complex industries.

GitHub Copilot is still the baseline tool that just works. It’s widely integrated, low friction, and reliable for everyday coding. Even if some of the newer tools are more advanced in certain areas, Copilot is often the one that keeps you moving without thinking about it. It’s not trying to redefine your workflow, it just quietly speeds it up.

The Playing Field

If we take a step back, OpenClaw deserves credit because it proved autonomous, local development is possible and useful, especially for developers who care about control and local execution. However, these six and dozens of other platforms coming online will take the local AI movement to the next level.

Leave a comment

By Mitch Stoller, Chairman of Literate AI

For fifty years, we lived in a world of “snobby” search. If you wanted to be discovered, you had to bow to the high priests of PageRank—the legacy media outlets like the New York Times or the Wall Street Journal. We spent our lives chasing the 2.6% of citations that come from Tier 1 publications, hoping a “blue link” would save our brand.

That world is gone. We are now in a “Zero Click” reality where an estimated 60% of searches end inside an AI-generated answer. Today, your most important audience isn’t a distracted consumer or a cynical editor—it’s a predictive algorithm.

As I’ve said before, Search is Snobby, but AI is Hungry. These Large Language Models (LLMs) are chomping at the bit for deep, technical, and specific information to answer the increasingly granular questions of your customers. To win in this environment, you need to stop writing for the “hook” and start writing for the “robots”.

That’s why we developed the R.A.D.D. framework. It’s the playbook for any brand or leader who wants to stop being invisible and start being the authority.

Relevance: Aligning with the Vector

Machines don’t read like we do. They break text into tokens and analyze mathematical patterns. To be Relevant, your content must align with the specific word patterns and “vectors” used by predictive algorithms.

Stop burying the lead. If you want to be the answer to a question about “portfolio risk in a volatile market,” you need to use the specific terms and positive sentiment associated with that solution immediately. If the machine can’t find a clear pattern, it will move on to someone else who made their points more clearly.

Authority: Owning the Owned

Your boring corporate website is now your most valuable digital asset. Why? Because you own it. In the age of AI, the machines look for sites with demonstrated authority based on links and depth.

Don’t wait for a journalist to validate you. By publishing high-quality, detailed content on your own domain, you establish yourself as the “Mount Olympus” for your specific niche. If you provide the most authoritative data, the LLMs will have no choice but to cite you.

Discovery: Optimizing the Bot’s Journey

The bots follow a specific journey: Relevance, Authority, Discovery, and Depth. To ensure Discovery, you must optimize your technical infrastructure.

LLMs prefer clean HTML over clunky PDFs. They need to find your pages easily to index them. If your “raw ore” is locked away in a format the machine can’t parse, you are effectively invisible to the very engines that are telling people what matters.

Depth: Answering the Next Question

Journalism is often shallow because it needs a click. AI is different; it wants to answer the follow-up. To achieve Depth, you must map out the 40 to 60 questions that define your category and answer every single one of them with meaningful detail.

If you provide a shallow answer, the user (and the bot) will look elsewhere. If you provide the depth, you stay in the “answer loop” and guide the consumer through their entire research journey.

The Golden Moment

We are in a golden window for Artificial Intelligence Optimization (AIO). The algorithms are still learning, and they are hungry for your expertise. But this window will close as platforms integrate advertising and become more selective.

Don’t just automate—Orchestrate. Use R.A.D.D. to turn your brand into the definitive source of truth. The robots are reading. Make sure they’re reading you.

RADD Scorecard

Leave a comment

To establish yourself as a category authority in the “Zero-Click” world, you must evaluate your content through the eyes of ChatGPT, Gemini, Perplexity, and the other LLMs. This R.A.D.D. Scorecard allows you to audit your existing articles and determine if they are feeding the hungry AI or simply whistling in the digital wind.

Welcome back to the 10 laws of AI Optimization.

Today we lay out our final law–the law to rule all the others. The foundation of marketing and companies in the AI age.

Orchestrate, don’t dictate.

Let’s explain what this means in two parts.

First, stop thinking in silos. That’s the big change we talked about in our last law–the Winner Take All Age is Over.

Here’s why. We have too many silos. Way too many. With different specialists handling each one.

Meta

AdWords

SEO

Content

Email

Comms

SMS

Linear TV

Streaming TV

YouTube

TikTok

AEO

Ecomm

Amazon

Walmart

Influencer

Affiliate

Who knows what else.

It’s just too much, too many different disciplines. Each is best for certain campaigns, messages and products. Each filled with subtilities and nuances. Each has a good or even great manager.

It’s very, very hard to manage across all these lanes.

Yet that’s about to change.

With the right AI tools, you can use an existing platform or build your own on top of an LLM, you will be able to coordinate across all these different channels.

Real-time, data-driven, syncing of customer actions across all your touchpoints.

Done right, you will be able to create a cohesive, personalized journey.

We’re not talking about marketing automation–the efficient execution of linear tasks. Like a series of emails.

Orchestration will be dynamic, context aware and cover all the bases.

Automation is doing things right.

Orchestration is concerned with doing the right thing at the right time for the right customer.

It’s the ultimate skill in AIO. True, it’s just in its early days. Heck, so is AIO.

But now is the time to start thinking and learning about where to start in AIO–making sure LLMs get your right–to where you can truly take. Orchestrated AI-driven brand and reputation-building and commerce.

McKinsey projects the B2C market alone could reach $1 trillion by 2030. That’s just in the US.

AIO demands a different kind of thinking–a holistic point of view matched with a high level of experimentation and flexibility.

Nothing sums it up better than our final rule.

Orchestrate, don’t dictate.

Leave a comment

I’m consistently surprised by how few clients have even considered the possibility of selling their products or services through ChatGPT.

Most are only just catching up to the idea that their customers are using it to research and compare providers. That alone still feels new. Teams are beginning to ask how they show up, how they influence answers, how they avoid being invisible.

But the next step rarely comes up. Why?

• No clear owner - It cuts across SEO, content, and sales, so no team takes responsibility.

• No familiar metrics - No traffic, no attribution, no dashboard, so it feels invisible to leadership.

• It requires mastering a new channel - New behaviours, new mechanics, and no established playbook.

• It disrupts what already works - Investing here forces uncomfortable questions about paid search, websites, and existing funnels.

So, what happens when that same interface moves from recommendation to transaction?

We are not fully there yet. The merchant infrastructure is early. The rollout is uneven. Most businesses cannot plug in overnight and start generating revenue.

But the direction is clear.

OpenAI has already signalled where this is heading. ChatGPT is evolving from a discovery layer into something closer to a full commercial environment. Consumers are already comfortable asking complex questions, comparing providers, and forming decisions inside a single interface. That behaviour is established. The next logical step is enabling those decisions to convert directly within the interface.

LLM product queries have already been determined as higher intent than conventional search. Soon the journey will compress from intent to purchase in a single thread.

That shift matters, even in its early stages, because by the time it feels “ready,” the rules will already be set. The same pattern played out with search, marketplaces, and social platforms. Early movers shaped visibility. Late adopters optimized within constraints they did not define.

Right now, most companies are still treating ChatGPT as a research tool. Something upstream of the real commercial moment. That framing underestimates what is happening. Once a system both recommends and enables action, it becomes a distribution channel.

There will be no switch from “not a channel” to “now a channel.” There will be a gradual increase in capability, followed by a rapid change in behaviour. By then, preparation will look like hindsight.

So the question now is less “should we sell through ChatGPT today?” For most, the answer is not yet.

The better question is: what would it take to be ready?

That means thinking about how your services are structured, how your offering is described, and how a model would interpret and recommend you. It means considering integrations, data exposure, and how you participate in a system where you do not control the interface.

Most businesses have not started that work…But they will need to.

Because ChatGPT is on a path to becoming a meaningful sales channel. And like every channel before it, the advantage will go to those who prepare before it is fully formed.

The window is open now.

For the better part of two decades, technology told a simple story: intermediaries were dying. The internet would connect businesses and consumers directly, remove friction, and flatten entire value chains.

And for a while, it did.

E-commerce enabled brands to go direct. Search engines gave consumers immediate access to information. Marketplaces streamlined distribution. Layers of brokers, agents, and gatekeepers began to disappear. Control shifted toward those who could capture attention and convert demand.

Disintermediation became the default assumption.

But that era is ending.

Large language models are a structural shift in how decisions get made. Search presented options, while LLMs synthesizes them. Platforms enabled discovery, while LLMs increasingly determine outcomes.

They aggregate information, interpret intent, and recommend answers.

And in doing so, they reinsert themselves directly into the transaction. This is a concept that I have dubbed “reintermediation”.

Disintermediation saw the removal of intermediaries (middlemen like wholesalers or agents) from a supply chain, allowing producers to sell directly to consumers, often via the internet/google.

Reintermediation is the reintroduction of new, digital intermediaries like LLMs to aggregate, curate, or simplify the buying process in a crowded market.

AI systems are becoming the new layer between supply and demand. Consumers no longer browse ten websites. They ask ChatGPT. Procurement teams do not evaluate dozens of vendors. They receive a shortlist. Increasingly, the journey collapses into a single interaction, mediated by an artificial intelligence that curates, filters, and decides.

The interface actively curates, ranks, and shapes outcomes.

It shows options and influences which ones matter.

That shift carries real consequences. Power moves away from those who own the product or the customer relationship and toward those who influence the model. Visibility is no longer just about ranking highly in search results or building brand awareness. Visibility depends on being selected, summarized, and recommended by systems that operate one layer above the open web.

In a disintermediated world, success depended on access. In a reintermediated one, it depends on inclusion.

This has immediate implications for both B2C and B2B models. Direct-to-consumer becomes mediated when discovery sits inside an AI layer. Brand equity functions as a signal within a model’s decision process. Even pricing power can shift as AI systems standardize comparisons and compress differentiation.

The battleground is moving.

It centers on reaching the systems that shape customer decisions.

That raises a harder question. If LLMs are the new intermediaries, who controls them? And who gets surfaced within them?

We are already seeing the early contours of this shift. Companies are beginning to think about AI optimization in the same way they once thought about SEO. The focus shifts to appearing in answers and being chosen within them.

The irony is hard to miss.

We spent twenty years trying to remove the middlemen. Now we are replacing them with something far more powerful and influential. I suppose we didn’t want to do our own research after all…

Leave a comment

When it comes to discovering information, Google has dominated the last 20 plus years.

Only with the 2023 launch of ChatGPT did the search monopoly begin to crack.

Now it seems to be such common knowledge even judges in Delaware acknowledge that search is no longer a monopoly—thwarting the breakup sought by the DOJ.

Of course, some folks are saying Google is going to win again.

From far behind in 2023, Gemini chatbot is now a leader, and Google search now often serves up AI Overviews instead of blue links. Plus just announced AI shopping tools.

Still, given Google’s cannibalizing its own core search business to stay relevant, it suggests it has got some serious competition this time around.

So what is going to happen?

Well, we may be done with the Winner Take All world.

Wouldn’t be such a bad thing.

Instead, we will have multiple sources for information and analysis, each with strengths and weaknesses. Instead of defaulting to Google.

Sure, it’s true that some of the LLMs, like ChatGPT, are far too reliant on single sources. Such as Wikipedia. But even that’s changing, driven by the LLM competition. Elon Musk’s Grok recently launched Grokipedia. Pretty rough now, but it will get better.

Plus we’ve got a full plate of LLMs with real user bases. Perplexity for research, Claude for coding and no doubt more custom LLMs coming soon to disrupt every disruptible field in the US and the world. Which is just about everything in professional services, the arts and information technology.

What does this mean for your brand’s or business’ efforts to optimize for AI discovery, reputation and ultimately brand awareness and sales?

First, it means that silos are over. Google’s total dominance of search and keywords is over. Meta’s dominance of digital product advertising is next. As the Chatbots play a bigger and bigger role in recommendations and discovery.

Second, it means you need to coordinate much more effectively across all your channels to succeed in a world where no channel dominates.

Turn into our final law to learn how to make it happen.

Leave a comment

We’ve all heard lots of things about the consumer journey, and why it’s important. But what is it?

Well, the idea was born as far back as 1898 by someone named Elias St. Elmo Lewis. They don’t make names like that anymore. An advertising pioneer, Lewis wanted to bring scientific rigor to selling products, and in those early days of industrializing, the challenge wasn’t building a brand or relentless competition. It was consumer ignorance.

New products were flooding the market, and the public needed to be educated before they could be persuaded.

Lewis developed a model to guide a prospect to a sale. The consumer journey, in other words. You’re probably familiar with Lewis’ model: AIDA (Attention, Interest, Desire, Action). As fit the media market of those days, it was linear.

• Attention: The first crucial step. If you fail here, Lewis argued, nothing else matters.

• Interest: Transform fleeting awareness into deeper curiosity about the product’s features, utility, etc.

• Desire: Go beyond the rational mind (”I get it”) and tap the power of emotions (”I want it”). It’s where brand building starts.

• Action: If all the other stages are done right, Lewis felt consumer action was a natural, inevitable consequence.

Today, more than a century later, the AIDA model is still in marketing textbooks. It aligns with behavioral psychology. It’s core.

Until today.

The chatbots blew it up.

Perplexity, Claude, Gemini, they don’t take the consumer journey. They’re not consumers. They’re predictive algorithms. They follow the beat of their own drummer.

They travel on the Bot’s Journey.

It’s not about Attention or Interest. It’s certainly not about desire. And the chatbots can’t actually do anything but, well, what they do. Analyze language and find patterns in response to queries.

So to map the bot’s journey, we need to know where they look for information and in what sequences. In other words, what happens after the first question is asked. So here’s a shot at the new AIDA.

Relevance–the bots crawl the web looking for pertinent information based on the analysis of word patterns.

Authority–the bots look for sites with authority based on links and traffic (we think). But it doesn’t work like Google sera

They select the best sources available. That doesn’t mean the New York Times. Behind a paywall. But too often it does mean Wikipedia.

Discovery–They prefer HTML to PDFs. All technical choices must be optimized to ensure the answer engines find the right material.

Depth–They favor information that responds as clearly as possible to the initial query and follow ups.

So, if you want to make sure you’re part of the Bot’s Journey.

It’s time to get RADD!

Leave a comment

For decades, the communications and marketing world has operated under a single, unchallenged theology: The “Mount Olympus” of credibility. If you landed a feature in the New York Times, the Financial Times, or the Wall Street Journal, you had reached the pinnacle. You weren’t just getting a story; you were buying a permanent seat at the top of Google’s PageRank system. Search was and is snobby. It favors the professional creative classes and the legacy media giants, leaving smaller players and even major corporate brands to fight for the scraps.

But the ground just shifted beneath our feet. Per a Britopian analysis, earned media accounts for a low of 5.7% to high of 15.4% of media citations in responses to queries related to the broad technology trends including cloud infrastructure, AI/data analytics, security/compliance and collaboration productivity.

Now drill down further. Within AI answers, the competition for “top-tier” visibility is intense. Per a Gemini analysis drawing on disparate sources, AI platforms often concentrate 34% of their journalistic citations on just one or two dominant brands.

Bottom line: we are spending millions of dollars and thousands of man-hours chasing a sliver of the digital pie that represents less than 16% of how the new world discovers information.

Summary of Citation Metrics for the

“Big Three”

Why is this happening? Because AI is hungry. Unlike the snot-nosed search engines of the past that cared about the prestige of a backlink, LLMs like ChatGPT, Perplexity, and Claude are optimized to provide specific, detailed answers to increasingly granular questions. Journalism, by its very nature and business model, cannot provide this level of deep, technical knowledge. A journalist needs a “hook” and a “narrative” to drive clicks; a robot just needs the facts to fulfill a query.

This is the reality of the “Zero Click” world. When 60% of searches end without a click because the answer is served directly by the AI, the “Blue Link” economy is effectively over. If you are still measuring success by your PR firm’s ability to get you into the Associated Press, you are measuring a ghost. The real battle for influence is happening in the other 97.4%—the technical white papers, the deep-dive blogs, and the structured data that the hungry machines are actually eating.

It is time to pivot from “prestige” to “presence.” You don’t need a journalist to validate your existence anymore; you need to feed the bot.

Leave a comment

Gemini sources for Citation Metrics for the “Big Three”

• https://arxiv.org/html/2507.05301v1

• https://ppc.land/chatgpt-referrals-surge-as-google-searches-hit-zero-click-records/

• https://www.position.digital/blog/ai-seo-statistics/

• https://ahrefs.com/blog/ai-seo-statistics/