Is generative engine optimization legit or just hype?

GEO is not a scam, but most of what's sold under the name isn't supported. The useful part is narrow and mechanical. The oversold part is the content-rewriting playbook.

The reasoning comes down to stages. An AI answer passes through four:

• Parametric: what the model already knows from training, with no web search.
• Access: whether a crawler can fetch and store your page.
• Retrieval: whether your page enters the model's context for a given query.
• Generation: which retrieved pages the model uses and cites.

Nearly every published GEO effect size was measured at the last stage, on pages already inside the model's context. Buyers want the earlier stages: getting found at all.

So when someone says GEO works, ask which stage they mean. For access, the answer is yes, and the engines document the mechanisms themselves. For rewriting pages to get chosen, the best evidence says no. For the longer version, read GEO explained honestly.

The critical survey of 45 GEO studies (arXiv:2607.14035, published 15 July 2026) states it plainly: "already-retrieved content can causally alter its citation or use, but no reviewed technique shows a stable, longitudinal, cross-platform causal effect on organic discoverability or downstream behavior."

Two other studies sharpen the point:

• C-SEO Bench (NeurIPS 2025) found most conversational SEO methods "are not only largely ineffective but also frequently have a negative impact on document ranking." Its stronger arm worked by moving the target document to position 1 in the context window, which no publisher can do.
• The original GEO paper (arXiv:2311.09735) is the source of the "up to 40% more visibility" claim. It tested an engine the authors built, with sources supplied in context, using a metric that counts words the intervention itself added. The gap between lab studies and live engines explains why that number doesn't transfer.

One caveat: the survey is single-author and not peer reviewed. Its conclusion still matches the better-designed experiments.

The levers with evidence remove blocks. They don't add a boost.

• Server-side rendering. Vercel found ChatGPT and Claude crawlers fetch JavaScript files but don't execute them. Content that only appears after JavaScript runs can be invisible to them. Googlebot renders, so this matters most for non-Google engines.
• Per-bot robots.txt permission. OpenAI says each crawler setting is independent. GPTBot governs training. OAI-SearchBot controls whether you appear in ChatGPT search answers. A blanket AI block can quietly remove you from ChatGPT search.
• Network reachability. A firewall or CDN can turn away a crawler that robots.txt allows. OpenAI's help page says eligibility needs both.
• Snippet and index controls. noindex, nosnippet, max-snippet:0 and data-nosnippet keep content out of AI answers. Templates sometimes apply them by accident.
• Specific, accurate facts. A price, a number or a one-line definition on a fetchable page gives an engine a sentence it can quote. It also corrects the facts AI platforms get wrong about specific brands.

None of these guarantees a citation. They decide whether you're in the candidate pool at all. The access fixes versus rewrites comparison puts the evidence side by side.

• Schema or FAQ schema for citations. Google says structured data isn't required for generative AI search. Ahrefs' matched test of 1,885 pages found AI Overviews citations fell 4.6% after adding schema. Existing markup does no harm, but it isn't a lever.
• llms.txt. Google says Search ignores it. No other engine documents reading it.
• AI-friendly writing and chunking. Google's AI optimization guide says you don't need to write in a specific way for generative AI search, and there's no ideal page length.
• E-E-A-T scores and domain authority. Google says E-E-A-T isn't a specific ranking factor. Domain Authority is a Moz metric no engine computes.
• AI rank tracking for Google. As of October 2026, Google's AI report exposes impressions only, so no position number can be checked.

These claims are easy to spot once you know the pattern. The list of GEO red flags covers the rest.

• Schema's untested corner. Ahrefs only studied pages already cited 100 or more times in AI Overviews. Whether schema helps a page enter the pool is untested, not refuted.
• Brand size. Well-known brands dominate citations, and no tactic controls for that. Being named on the sites engines already cite has a strong association with appearing in answers, with no promised effect.
• Answers without search. Retrieval fires per query, at the engine's discretion. When it doesn't, the model answers from training data, and nothing you publish this quarter reaches that answer.
• Injected content. Hidden text and prompt injection aimed at AI agents can change outputs in lab tests. It's illegitimate and a reputational risk, not a tactic.

1. Check robots.txt for each retrieval bot: OAI-SearchBot, PerplexityBot, Claude-SearchBot and Googlebot.
2. Request your key pages as each AI crawler and confirm the firewall returns 200, not 403.
3. Compare raw HTML with the rendered page. Move prices, product names and definitions into the server response.
4. Search templates for stray noindex or nosnippet.
5. Ask AI engines your buyer questions, note the wrong facts, and state the correct ones on pages crawlers can fetch.

For step 2, the AI Crawler Checker requests a page as each crawler and shows which ones get turned away. Before you pay anyone, test them against which GEO tools are worth paying for.