How AI Sees You

Why each AI platform needs a different approach

Updated September 22, 2026

AEO (Answer Engine Optimization) and GEO (Generative Engine Optimization) are the terms for structuring content so AI tools select and cite it. The two overlap heavily, and both build on top of solid traditional SEO rather than replacing it. What differs is how each platform actually sources its answers, which is worth understanding before assuming a single approach works everywhere.

ChatGPT: selective, and reads more of the page than you think

ChatGPT doesn't crawl the web live. For questions needing current information, it breaks the question into narrower sub-queries, runs those against a search index, and reads more of the actual page than a search snippet would show. It tends to cite fewer, stronger sources per answer rather than everything remotely relevant, so ranking as one of those few sources is worth real effort.

"Authoritative" here means topical depth more than domain age or backlinks: does a site have several related pages on a topic, or is this a single post covering it once among a hundred unrelated things. Answer the specific question in the first sentence of a section, use a heading that states the topic plainly, and include the kind of specific detail, a real number, a named tool, an exact step, that only someone who actually did the thing would know.

Check this first: none of it matters if GPTBot is blocked. Check Robots.txt for a rule disallowing GPTBot specifically, including older, broader bot rules that might catch it by accident.

Perplexity: citation-dense, and rewards specificity over authority

Perplexity is built as a research tool first. It cites more sources per answer than ChatGPT, leans heavily on real-time web data over a static index, and is comfortable pulling from community sources like Reddit when they carry detail a broader source wouldn't include. This is genuinely good news for smaller sites: specificity beats domain-wide authority here.

Publish and update on a visible schedule, since a real modified date signals the freshness Perplexity weighs. Go narrow before broad: a page specifically about one exact problem often outperforms a broader page covering it as one section among several. Include the exact setting, number, or step, and don't worry about sounding overly formal.

Check Robots.txt for a rule blocking PerplexityBot specifically before assuming a content quality problem.

Google AI Overviews and AI Mode: tied directly to your existing rankings

Of the major platforms, Google is the least separate from work you're likely already doing. As of July 10, 2026, AI Mode became the default way Google answers most searches, running on the Gemini 3.5 Flash model, with the classic ten blue links pushed further down the page rather than removed. It still draws from the same index Google uses for regular results, so a page needs to already be a strong candidate in traditional search before it has a real shot at being cited. Technical health, page speed, and backlinks still function as gatekeeping signals, the same as for normal rankings.

AI Mode also works differently under the hood: it often splits one query into several related sub-questions (query fan-out), answers each separately, then merges the results. A page that only covers the exact phrase someone might search for can lose out to a page that also answers the two or three questions people naturally ask alongside it.

On top of that foundation, front-loaded direct answers and clear heading structure raise the odds of being the page an answer quotes. FAQ schema itself is no longer part of that equation for Google specifically: Google retired the FAQ rich result in May 2026, though the schema is still harmless to keep and can help other AI tools parse the page.

Google also expanded Preferred Sources into AI Overviews and AI Mode in May 2026. A signed-in user can manually star a site they trust at google.com/preferences/source; from then on, that site's links get a visible "Preferred" label and are shown more often to that user. There's no schema or code to add, since it's entirely reader-driven - the only real lever is asking your regular visitors, subscribers or customers to add you.

If a page isn't already ranking on page one for its target query, that's still the fix to prioritize before layering on any AI-specific tactic.

Claude: the most conservative about reaching for outside sources

Claude answers most questions from its training data alone and only searches the web when the question needs current or verifiable information, comparisons, recommendations, and other discovery-style questions are far more likely to trigger a search and a citation than plain informational ones. When it does cite, it favors a small, curated set of sources written in a balanced, non-promotional tone over anything that reads like marketing.

Anthropic runs three separate bots, and a rule for one has no effect on the others: ClaudeBot (training), Claude-SearchBot (search indexing), and Claude-User (fetching a page a person asks about directly). If the goal is showing up in Claude's answers, Claude-SearchBot and Claude-User are the two that matter, and each needs its own Robots.txt directive.

For product and service questions, Claude tends to cite the independent reviewer over the product's own page, worth factoring in for review or comparison content specifically.

Gemini: grounded in live Google Search, plus real strength with images

Gemini works differently from the others: when it needs current information, it runs a live Google Search behind the scenes, a process Google calls grounding, and builds its answer from whatever comes back. That means the single biggest lever isn't writing for Gemini specifically, it's ranking well in regular Google Search. If a page isn't on the first page or two for the query, Gemini is unlikely to see it during grounding at all.

On top of that, Gemini leans more heavily on Schema Markup and reads images and video more reliably than the other platforms, so a clearly labeled screenshot or diagram can earn a citation on its own.

Check Robots.txt for both Googlebot and Google-Extended, a separate token controlling whether Google's AI products can use the content. A site can rank normally in Google Search while still blocking Google-Extended, quietly excluding it from Gemini with nothing looking broken on the surface.