Mike King vs Google is the only fight that matters

Mike King published two pieces this week and one last weekend that, taken together, form the most pointed critique of Google's GEO posture anyone in the industry has put on paper. The Personal Intelligence experiment. The agentic RAG breakdown. And the direct accusation that Google's recent generative AI guidance is "naive and self-serving."

The response from the rest of the industry has been roughly: nod, share, move on.

That's the wrong response. King is doing something specific here, and the SEO industry is treating it like just another thought-leadership cycle. It isn't. He's running real experiments, publishing data with confidence intervals, and pointing out — with receipts — that the platform most consultants still treat as the arbiter of truth has a structural conflict of interest about what it tells the industry.

If you're a UK business owner trying to figure out who to believe about AI search right now, this is the fight to watch. Because the outcome decides what advice you're going to be paying for in twelve months.

What King actually published this week

Three pieces. Worth understanding what each one claims before you read the take.

The Personal Intelligence study tested whether seeding brand signals into a Gmail account changed which brands AI Mode recommended. Across 1,922 responses, seeded brands appeared 46 percentage points more often in the Personal Intelligence-connected account than the control. Gmail-based signals were dramatically stronger than Photos. Brands also moved into better positions — +23 points in top-3 placement, +43 points in top-10.

The agentic RAG piece argued that every major AI search platform — Google AI Mode, ChatGPT Search, Perplexity, Claude, Gemini Deep Research — has moved past single-shot retrieval. They plan, route between tools, retrieve iteratively, and grade their own drafts before answering. King's claim: if your GEO program is still optimised for one-shot retrieval, you're optimising for a system that no longer exists.

The Google guidance critique was the bluntest of the three. King points at Google's updated generative AI optimisation doc — the one that essentially declared AEO and GEO "still SEO," dismissed chunking, and waved off rewriting for AI — and reminds the industry that two years ago the Content Warehouse leak showed Google's public guidance and internal reality diverging by a lot. He contrasts Google's "trust us, keep doing what you were doing" posture with Bing publishing actual citation tooling and openly calling GEO an emerging discipline.

Three pieces. One argument underneath all of them: the public guidance from Google is calibrated to keep the industry doing janitorial work whilst the actual mechanics of AI search move somewhere Google's guidance won't tell you about.

Why this matters more than the usual Google-vs-SEOs cycle

We've been here before. Google publishes guidance. Some SEOs say "see, it's just fundamentals." Other SEOs say "they're lying." Twitter argues for a week. Nothing changes.

The industry's job is to notice when the official line and the working reality have separated. Right now they've separated badly.

What's different this time is the asymmetry of evidence. King isn't just disagreeing with Google's framing — he's publishing controlled experiments that contradict it, then naming specific architectural shifts (agentic retrieval, personal context signals) that Google's public docs simply do not address.

You can't credibly read the Personal Intelligence study and then tell a client "Gmail isn't a ranking signal, ignore it." You can't read the agentic RAG breakdown and then tell a client "AI search is just SEO with structured data." The evidence has outrun the official line.

The industry's job is to notice when the official line and the working reality have separated. Right now they've separated badly.

Google's incentive problem, stated plainly

I wrote about this in Mike King is right about Google. And right about Bing. a few weeks back, but King's latest piece sharpens the point.

Google's incentives in 2026 are not what they were in 2016. For two decades, Google's interest and the open web's interest were roughly aligned — Google needed publishers to produce indexable content, publishers needed Google to send traffic. Now the deal has changed. AI Overviews and AI Mode reduce the need to click. Competitive AI platforms are eating attention Google used to own exclusively. Referral traffic is shrinking. The leverage Google had to define what "good content" means is the weakest it has been in this century.

In that environment, "keep doing SEO, it's all you need" is not neutral advice. It's an attempt to keep an industry focused on optimising for Google whilst Google's share of discovery erodes. The advice isn't wrong, exactly — fundamentals do still matter — but it's incomplete in a direction that conveniently serves Google.

Bing's posture is different not because Microsoft is morally superior but because Microsoft is the challenger. Challengers publish tooling and tell publishers what's actually happening because they need publishers to participate. Incumbents publish guidance that minimises change because change is risk.

That's the loop. And we've been in it before — it's just the names have moved around.

What the SEO industry is doing wrong with this

Two failure modes, both common, both in evidence right now.

The "see, it's just SEO" camp is using Google's guidance as permission to not learn anything new. The fundamentals haven't changed, so why bother understanding agentic retrieval or personal context signals or knowledge graph grounding? This camp is going to be fine for about eighteen months and then suddenly very much not fine, because the surface they're optimising for is genuinely shifting and they've talked themselves out of noticing.

The "everything is changing" camp is using King's work as a sales tool for whatever GEO product they're selling this quarter. They've turned the agentic RAG piece into a slide deck and the Personal Intelligence study into a workshop and the Google guidance critique into a tweet thread, without acknowledging that none of King's pieces tell you what to do — they tell you what's happening. There's a difference.

The honest position is in the middle, and it's uncomfortable because it doesn't sell as well as either extreme. The fundamentals still drive AI citation. *And* the architecture has moved past single-shot retrieval. *And* personal context is becoming a measurable signal you can't audit. *And* Google's public guidance is a poor map for the territory. All four things are true simultaneously, and a serious consultant has to hold all four without flinching.

What this means if you're paying someone for SEO right now

If you're a UK business owner working with an agency or consultant on SEO or "AI search optimisation," there are three questions worth asking this week.

First: what does your consultant think of Mike King's Personal Intelligence study? Not "have they read it" — what do they think it means? If they haven't heard of it, that's a data point. If they have but dismiss it as US-only or not relevant to your business, that's a different data point. If they engage with it honestly — including the parts that are genuinely uncomfortable for the consulting industry, like the fact that Gmail signals are unauditable — that's the answer you want.

Second: do they distinguish between what Google publishes and what Google does? An agency that quotes Search Central guidance as if it were physics is an agency that hasn't read the Content Warehouse leak, or hasn't updated its mental model since 2019. Either way, the work is going to be calibrated to a Google that doesn't exist anymore.

Third: are they comfortable saying "we don't know"? Agentic retrieval is a black box. Personal context is invisible to publishers. AI citation tracking is genuinely immature. The honest answer to a lot of GEO questions right now is "we have hypotheses, we run tests, we adjust." Anyone selling certainty in this environment is selling something else.

The piece I'd actually like to see next

I'm not going to write King's next piece for him, but the missing piece in his trilogy is the one about measurement. He's done a brilliant job describing what's changed — architecture, personal context, the gap between Google's guidance and Google's behaviour. He hasn't yet written the piece that says: given all of this, what does an honest measurement framework look like for a business that doesn't have iPullRank's research budget?

Because that's where the consulting industry actually lives. Not at the SEO Week stage. In the meeting room where a managing director with £8,000 a month of marketing budget asks "is this working?" and a consultant has to give an answer that's both honest about the measurement limits and useful enough to justify the next invoice.

Right now, most of the industry is answering that question by pretending the measurement problem doesn't exist. Citation count, share of voice, AI Overview presence — pick a number, draw a chart, call it a quarter. King's work makes it harder to keep doing that with a straight face, which is exactly why the industry would prefer to nod, share, and move on.

Don't move on. The fight he's picking is the only one that matters this year.