Google's First AEO Guide Says Skip llms.txt, Chunking, and AI Schema
Google published its first official guide to optimizing for generative AI search on May 15, 2026, and explicitly framed AEO and GEO as repackaged SEO rather than separate disciplines. The mythbusting section names five tactics to ignore: llms.txt files, content chunking, AI-specific writing, inauthentic mentions, and special schema markup. That list covers most of the line items on a typical AEO retainer.
The guide and an accompanying blog post went live the same day. Matt G. Southern at Search Engine Journal picked it up within hours, and the SEO consultant Slack groups did the rest. By the time most agencies opened the link Friday afternoon, they were already drafting client emails.
The exact sentence vendors had been hoping Google wouldn't write
"From Google Search's perspective, optimizing for generative AI search is optimizing for the search experience, and thus still SEO."
That's the line. It defines AEO as "answer engine optimization" and GEO as "generative engine optimization" up front, so nobody can claim a misunderstanding of vocabulary. Then it folds both into SEO and walks straight into the mythbusting section.
The five mythbusts, with Google's wording lightly trimmed:
- llms.txt and AI text files: "You don't need to create new machine readable files, AI text files, markup, or Markdown to appear in generative AI search."
- Content chunking: "There's no requirement to break your content into tiny pieces for AI to better understand it." Google adds that its systems "are able to understand the nuance of multiple topics on a page."
- AI-specific writing style: "You don't need to write in a specific way just for generative AI search." Synonym farming and "AI-friendly" rewrites earn a direct dismissal.
- Inauthentic mentions: "Seeking inauthentic 'mentions' across the web isn't as helpful as it might seem." A polite way of saying brand-mention farms don't move anything Google ranks.
- Special schema: "Structured data isn't required for generative AI search, and there's no special schema.org markup you need to add." Schema is still useful, but not as an AI-visibility lever.
Each one of those is the headline tactic in a published AEO playbook somewhere. The chunking advice in particular has been pitched as the central technical lift in AEO retainers since mid-2025.
Why a paragraph in developer docs is the most consequential thing this week
Google rarely publishes documentation that walks straight into an active paid service category and contradicts it. The closest recent analogue I can think of is the structured data limits page from 2023, and even that one waited until after the affected guidance had quietly died on its own.
This one didn't wait. AEO and GEO as service lines emerged roughly eighteen months ago, mostly on the back of decks claiming that LLMs index differently than Search. Some of that was always true for Perplexity and Claude. Almost none of it was true for Google's AI Overviews and AI Mode, which is what the average mid-market client was actually buying coverage for.
The industry response has been quick and predictable. Market My Market put numbers on it, citing $2,500/mo AEO line items added to existing $5,000 SEO retainers, a 50% revenue lift on the same client. Several agencies have already started rebranding AEO and GEO as "AI visibility strategy" or "LLM citation management," same playbook, new invoice line. The relabeling will probably work for a quarter, maybe two, before clients start asking the same question the Ahrefs 1,885-page schema study forced last quarter: where's the lift.
What the doc isn't saying
Three carve-outs worth flagging before anyone reads the guide as permission to delete a category.
First, this is Google's surface only. ChatGPT, Perplexity, and Claude still pull from a different mix of sources and weight them differently. The Kevin Indig consensus study put cross-engine citation overlap at 2.37%, which means optimization for the non-Google surfaces is still a separate problem with separate evidence. Google's doc doesn't pretend otherwise.
Second, Google explicitly says structured data is still valuable for regular Search and recommends continuing to use it. The mythbust is narrower than the rebrand crowd is making it sound. The claim is that schema isn't a special AI lever, not that schema doesn't matter.
Third, the doc doesn't argue that the work is unimportant. It argues that the framing is. Quality, originality, technical fundamentals, and freshness still drive AI Overview visibility. Those are the same things that drive Search visibility. The doc's preferred phrase is "non-commodity content," which seems to mean roughly the same thing as everyone else's "people-first" language with one additional emphasis: stop publishing the obvious thing.
The three questions I'd send any AEO vendor on Monday
If you're carrying an AEO or GEO retainer right now, the cleanest audit move this week is to map the line items against Google's mythbust list and ask the vendor to defend each one with attribution. Three questions are usually enough to do most of the filtering:
- Show me Google AI Overview citation lift you can attribute to llms.txt or any AI-specific file you've shipped on our site.
- Show me a citation gain in AI Overviews or AI Mode tied to schema you added specifically because it was "AI-friendly," not part of a normal Search rollout.
- Show me Search Console impression lift from content chunking work, separate from rewrite or freshness work, in the last 90 days.
If the answer is "we don't track Google AI Overview attribution at that resolution," you've found your line item. That's not a vendor failure on its own; the attribution genuinely is hard right now. But the retainer was sold on the premise that the work would move that needle, and the lift either exists in data or it doesn't.
For non-Google surfaces, the same vendor might still be worth keeping. The Omniscient brand-citation study and the Microsoft Clarity AI citation tab both pointed at work that has a real signal, just not under the same labels.
The line item I'd kill first
If I had to pick one of the five mythbusts to act on this week, it would be llms.txt. It's the most expensive tactic relative to evidence, the cleanest to attribute (the file either exists or it doesn't), and the easiest to defend killing in a budget conversation. Google just gave you a documented sentence to point at.
I don't think this guide ends AI search optimization as a discipline. It probably just ends a particular sales pitch that's been running on the assumption Google would never write the sentence it just wrote. From what I've seen, the practitioners who were already doing this work as "SEO with a few new endpoints to monitor" just got told they were right. Everyone selling it as a separate retainer has a much harder Monday morning.
Notice Me Senpai Editorial
