Is AI SEO a Scam? An Operator's Answer

Skeptics asking this question have good instincts: the loudest corner of marketing right now is people selling "AI SEO" with statistics nobody can trace. We build and operate production authority sites, so we have no need to defend this industry, and some standing to criticize it. The honest answer splits cleanly in two, and since May 2026 Google has said so in writing.

Is AI SEO a scam?

No, not as a practice: structuring pages so AI answer engines can retrieve, quote and cite them is real work, and Google's guide to optimizing for generative AI features, published May 15, 2026, puts it plainly: "optimizing for generative AI search is optimizing for the search experience, and thus still SEO" (Google). What is often a scam is the sales layer wrapped around that work.

Google is blunt about that layer too. Its guidance on third-party SEO services, updated June 5, 2026, lists "AEO" and "GEO" tools by name among the services to think critically about, says "Google doesn't evaluate third-party services, so be wary of such claims," and adds that third-party tools "don't have access to our internal ranking data. They can't guarantee performance" (Google). Its test for good advice is one we would sign: it "either qualifies their claims as opinion based on data or experience, or backs up their claims by citing official Google Search guidance."

What does Google say AI SEO vendors get wrong?

Google's May 2026 guide includes a "mythbusting" section that rules out most of what gets sold as an AI SEO deliverable. In the table, llms.txt means the llms.txt proposal (llmstxt.org): not an adopted standard, and no major engine has committed to reading it. The pitches we see most often, checked against Google's own documentation as read on 2026-10-14:

Google is one engine, and this table speaks for Google Search only. It still settles the most common pitches as far as Google goes: the "special AI files and markup" deliverables do nothing there, by the operator's own account.

What part of AI SEO is real?

The core practice is real and measurable: structuring pages so AI answer engines, including ChatGPT, Perplexity, Google AI Overviews and AI Mode, Copilot and Gemini, can retrieve, quote and cite them. The discipline even has a peer-reviewed origin. The Princeton-led GEO study (KDD 2024) coined the term generative engine optimization, GEO, and reported that its methods boosted visibility by up to 40% in a benchmark of generative engine responses, for tactics like adding source citations, quotations and statistics.

What Google asks for is ordinary, demanding work: its 2026 guide says unique, non-commodity content will likely influence a site's presence in generative AI search "more than any of the other suggestions in this guide," and that AI features rest on its core ranking and quality systems. The same "no special markup" point appears on Google's AI features page, while its May 2025 guidance repeats the people-first fundamentals it has published for years. The demand side is documented too: Semrush's 17-month clickstream analysis of ChatGPT tracks real referral behavior from AI assistants to websites.

In our own fleet, the practice produces observable events, not just theory: AI crawlers fetch our pages routinely in the server logs, our GA4 property has a dedicated AI-assistant referral channel receiving traffic, and a new glossary page earned a documented Google AI Overview citation within days of shipping [our data]. The definitional groundwork is in what GEO is, and the complete operating detail lives in our generative engine optimization guide.

What does the evidence actually support?

The strongest available evidence supports structural tactics, and punctures most sales pitches, often in the same study. Here is the claim-by-claim ledger.

Read the table twice and the pattern shows. The tactics with evidence are unglamorous structure work (Ahrefs' 1,885-page schema test and the Princeton benchmarks agree on that), while the numbers used to close sales are conversion multipliers with survivorship bias built in, or pure inventions. Click studies cut against everyone, us included: Ahrefs' CTR analysis and Pew's browsing data both show AI summaries reducing clicks.

Where does the scam start?

The scam starts exactly where claims detach from checkable sources. Three patterns account for most of the snake oil.

Laundered statistics. A number appears in one vendor blog without methodology, gets cited by a second vendor, and by the third hop it reads like established science. The canonical case is "Perplexity's 0.75 content quality score," repeated across dozens of posts and published by no primary source, ever. If a figure cannot be traced to a named study or platform document, it is not evidence; it is decoration.

Outcome promises. No vendor operates the answer engines, so no vendor can commit an engine to citing you. Google's own AI-features guidance describes eligibility and best practice and states that indexing and serving aren't guaranteed, even for pages that meet every requirement. A contract that promises specific citations, rankings, or traffic is a promise about a system the seller does not control.

Hype outrunning adoption. The llms.txt proposal (llmstxt.org), not an adopted standard, and no major engine has committed to reading it, is routinely sold as a mandatory AI-SEO deliverable. Google's 2026 guide now says Google Search ignores llms.txt files, so creating one "will neither harm nor help" visibility there. Google's John Mueller had earlier compared it to the keywords meta tag, and Search Engine Land published a reasonable counterpoint for other systems that might use the file. We serve it across our fleet and count the bot requests ourselves; the running log is in our llms.txt watch, and our llms.txt guide shows how to write the file yourself rather than buy it as a deliverable.

How do you tell an honest operator from a snake-oil seller?

Five checks separate them, and each takes under a minute.

1. Trace one statistic. Ask where their headline number comes from. An honest operator names the study and its date; a seller says "industry data."
2. Ask what failed. Anyone actually running experiments has negative results. If every tactic they tried "worked," none of it was measured.
3. Ask for receipts. Server logs, Search Console screenshots, before-and-after data from properties they operate. Commentary about other people's case studies is not operating experience.
4. Refuse promised outcomes. The moment rankings, citations, or traffic are promised, the conversation is over. Nobody can promise this, and Google says outright that third-party tools can't guarantee performance.
5. Follow the incentive. If every blog post concludes that you need their tracking tool or retainer, you are reading a funnel, not research.

Should you pay anyone for AI SEO right now?

Usually not yet, and we say that as a company that sells authority builds. Most sites can capture the evidence-backed portion of GEO with their existing team: restructure the highest-value pages answer-first, keep answers in raw HTML, allow the documented AI crawlers, and measure for 90 days before spending a dollar on tools or agencies. For Google, the measuring is free: its guide points site owners to the Generative AI performance report in Search Console.

Whether citations repay even that effort is a fair question. The click math, including Pew's ~1% citation-click finding, is worked through honestly in are AI citations worth anything. If the 90-day measurement shows movement and your volume justifies help, the market's real price ranges and what they buy are documented in what GEO agencies charge. Hire with sourcing standards in the contract, and keep the five checks above.