Meta is dying. Google just told you why.

On May 8th, Julia Angwin called Meta's quarter the start of a long zombie era. Daily active users fell from 3.58 billion to 3.56 billion. Record revenue, record costs, and an AI division that doesn't appear in the top 100 most-visited websites despite a $100 billion spend.

Two days later, Google quietly published a documentation page telling you that AEO and GEO are "still SEO."

Most people read those as separate stories. They aren't. They're the same story told from opposite ends — one company that knew what business it was in, and one that still doesn't.

The Levitt frame still works

Theodore Levitt's *Marketing Myopia* argument from 1960 is the cleanest lens for what just happened. Companies fail when they define their business too narrowly. Railroads thought they were in the railroad business, not transportation. Trolley operators thought they were in the trolley business, not movement. The product changed; the underlying job didn't.

Greg Jarboe at Search Engine Journal applied this to Meta this week and the read holds up. Six major pivots in 22 years. Social network, then mobile, then video, then metaverse, then generative AI. Reality Labs has burned roughly $80 billion. Meta.ai isn't in the top 100 globally. ChatGPT is at 5.7 billion monthly visits, Gemini grew 283.8% year-over-year, Claude grew 423.7%. Meta isn't on the chart.

Google, meanwhile, sits at 86.9 billion monthly visits. YouTube does 29.3 billion. The gap between Google and Facebook isn't a market share gap. It's the data signature of two different business definitions.

The business Google thinks it's in

Read the new optimisation guide Google published on the 15th and you'll see a company that has decided, on the record, what business it's in.

It isn't "the AI Overviews business." It isn't "the AI Mode business." It isn't even "the search results business" in the SERP-as-product sense the industry has spent twenty years optimising for. Google's documentation now states plainly that generative AI features are "rooted in our core Search ranking and quality systems" and that optimising for them "is optimising for the search experience, and thus still SEO."

That sentence is doing more work than people are giving it credit for.

It tells you Google has resisted the temptation to redefine itself around the surface — to become "the AI answer business" the way Meta tried to become "the metaverse business." It's still defining itself as the information access layer. The interface changed. The job didn't.

The interface changed. The job didn't.

That's why the same documentation tells you llms.txt isn't required, chunking isn't required, and there's "no special schema.org markup to add" for AI features. Not because those things are necessarily worthless — but because Google has decided its job is to surface useful content regardless of how publishers tag it, and to define itself by that job rather than by the tactics layered on top.

It's exactly the discipline Meta lacks.

Meta keeps redefining the product, not the job

Look at how Meta has talked about itself since 2021. Each pivot has been framed as a redefinition of *what we make*. Metaverse. Generative AI. Reality Labs hardware. Smart glasses. AI agents. Each time, the company has tried to convince the market that the product category is the strategy.

But the underlying job — the actual reason 3.5 billion people opened those apps — was never "I want a metaverse" or "I want a Llama-powered chatbot." It was the social and informational gravity that Facebook, Instagram and WhatsApp accumulated over twenty years. The job is "show me what people I care about are doing and saying." Meta has spent five years and well over a hundred billion dollars building products that answer different questions.

And the Q1 squeeze tells you what happens next. Ad impressions up 19%. Ad prices up 12%. Revenue per user up 27%. Daily users down. That's not a growth playbook. That's a yield-extraction playbook on an asset that's quietly shrinking. Levitt described this exact trap — companies focusing on selling the current product harder rather than understanding the underlying need.

What this means for the SEO conversation

Now stack the Google documentation against the Ahrefs report on schema markup, the FAQ rich result deprecation, and the GA4 AI assistant channel — three things that all dropped within the same week.

The Ahrefs study tested 1,885 pages and found schema didn't move AI citations. AI Overviews showed a 4.6% decline against controls. AI Mode and ChatGPT showed shifts indistinguishable from noise. FAQ rich results, which an estimated 168,000 pages were optimising for as "critical for GEO," have been removed entirely. GA4 has started natively classifying AI assistant traffic, which finally lets you measure what was previously invisible.

What these three things share, taken together, is a Google that's behaving like a company that knows what business it's in. The visible rewards for tactical markup are narrowing. The honest data is showing that the markup itself isn't the lever. The measurement infrastructure is being made boringly official rather than left to vendor cottage industries.

That isn't Google deprecating SEO. That's Google narrowing what SEO actually is, back toward the load-bearing parts.

The mistake the GEO industry made

A large part of the GEO industry made a Meta-shaped mistake over the last 18 months. It defined itself by the tactic rather than the job.

The pitch was: AI search is a new discipline, it requires new markup, new tools, new files, new dashboards, new acronyms. GEO. AEO. AIO. LEO. The implicit business definition was "we sell the AI optimisation product." And like Meta's metaverse pivot, that definition depends on the surface being durable enough to organise a discipline around.

It isn't. The Ahrefs data, the FAQ removal, and Google's documentation all point the same direction: the surface keeps moving, and the things that hold up are the things that always held up. Useful content. Real authority. Brand presence in the places real people look. Clean technical hygiene. The boring spine.

The consultants and agencies that defined themselves as "the search visibility business" — the job, not the tactic — are doing fine through this. The ones that defined themselves as "the GEO product" are in the Meta position. Selling harder against a shrinking surface.

What to do with this

If you run marketing for a business, the practical takeaway is uncomfortably simple. Audit the language your agency or in-house team uses to describe what they do. If it's defined around tactics — schema, llms.txt, chunking, AI citation tracking platforms — you have a Meta-shaped problem in miniature. The product they're selling you may be the wrong unit.

If it's defined around the job — being the most useful and visible answer to the questions your customers ask, wherever they ask them — you have a Google-shaped operation. The surface can change underneath them and the work still compounds.

This is the actual line the protocol layer shift is drawing. Not between SEO and GEO. Between people who define their work by what they optimise for, and people who define it by what they're trying to achieve. The first group spends the next five years renaming itself every eighteen months. The second group keeps shipping.

Levitt's essay was 66 years old when Meta started losing users and Google published its "still SEO" guide in the same week. Both events are footnotes to the same point. The companies and consultants that survive the next ten years will be the ones who can answer, in a single sentence, what business they're actually in.

The ones who can't are already in the zombie era. They just haven't noticed yet.