How do I optimize my website for ChatGPT, Gemini and Perplexity?

The three engines share a foundation and differ enough at the margins to matter. Roughly seventy percent of the work is common; the remaining thirty is where teams waste a quarter by applying one engine’s playbook to another.

The shared seventy percent

Do these once and they pay into every engine.

• Server-rendered content. The answer exists in raw HTML before any script runs.
• Allowed crawlers. In robots.txt and, more importantly, at the CDN or WAF.
• One complete answer per page, stated in the first two sentences under a question-led heading.
• Specific, attributable claims. Dated numbers, named sources, original data.
• Consistent entity description. One sentence stating your category, audience and alternatives, used identically everywhere.
• Structured data that matches the visible text.
• Ordinary SEO health. Indexed, internally linked, reasonably authoritative — because every engine retrieves before it selects.

If none of that is in place, engine-specific tactics are premature. Do the seventy percent first.

Where they differ

ChatGPT

The engine most of your buyers actually use, the hardest to move, and the slowest to respond.

What it rewards: consistent third-party description above almost everything else. It is common to find a competitor ranking below you on Google being named first by ChatGPT, purely because a dozen independent sources describe them the same way and describe you not at all.

The mechanical requirement: GPTBot does not execute JavaScript. An analysis by Vercel and MERJ covering more than 500 million GPTBot fetches recorded zero JavaScript execution — it downloaded JavaScript files roughly 11.5% of the time and never ran them. A client-side-rendered page shows it an empty shell regardless of how well it ranks.

What to prioritise: fix rendering first, because it is binary. Then work the corroboration map — the specific third-party domains ChatGPT cites when answering your category’s questions. Being accurately described on five of those beats fifty generic mentions.

Expect slow feedback. Several weeks minimum, because changes depend on recrawling and on third-party sources updating. Judging a programme on ChatGPT at week three will always look like failure.

Gemini and AI Overviews

The most SEO-dependent surfaces, and therefore the ones where your existing work already counts.

What they reward: classic organic strength. Candidates are drawn from Google’s index, so ranking well for a query or its close relatives is the main entry requirement. AI Overview presence has held steady at roughly 32.5% of monitored queries through 2026, in a narrow 31–34% band.

The mechanical difference: Googlebot renders JavaScript. The rendering problem that makes sites invisible to ChatGPT does not apply here in the same way — which is why a site can be cited in AI Overviews and absent from ChatGPT entirely, and why measuring only one engine produces confident wrong conclusions.

What to prioritise: get the page into the top twenty organically, then optimise at passage level. Identify the specific sub-question you could own most cleanly, answer it in two or three self-contained sentences under a question-led heading, and add one concrete fact.

Perplexity

The most citation-hungry of the three and by far the most useful for diagnosis.

What it rewards: retrievable, current, clearly-structured pages. It retrieves aggressively at answer time and cites openly and prominently, which makes it the fastest place to see whether a new page has been picked up.

The mechanical requirement: same as ChatGPT — PerplexityBot does not execute JavaScript.

What to prioritise: use it as an instrument rather than a target. Publish or update a page, then check Perplexity within a week. If it picks the page up, retrievability is confirmed and any remaining absence elsewhere is a selection or corroboration problem. If it does not, you have a mechanical fault and no amount of editorial work will help.

The trap: Perplexity’s responsiveness makes programmes look successful early. A quarter that moved Perplexity substantially and ChatGPT not at all is normal at six weeks and a genuine concern at six months.

Which to target first

Sequence by where your buyers are and what you can verify.

1. Fix the shared foundation. Rendering and crawler access, across all engines simultaneously. One fix, three benefits.
2. Verify with Perplexity. Fastest confirmation that content is retrievable.
3. Work Gemini and AI Overviews via SEO. Your existing organic programme is already most of this, so the marginal cost is passage-level editing rather than new work.
4. Then invest in ChatGPT corroboration. Slowest, hardest, highest commercial value, and the one that requires PR rather than marketing execution.

Running them in the reverse order — starting with ChatGPT corroboration before confirming your site is even readable — is common and expensive, because the outreach points at pages the crawler cannot read.

What transfers and what does not

Transfers completely: rendering fixes, crawler access, answer-first structure, specificity, entity clarity, schema accuracy. These are the seventy percent.

Transfers partially: organic ranking work. It is close to decisive for Gemini and AI Overviews and merely helpful for ChatGPT and Perplexity, which retrieve from their own pipelines.

Does not transfer: corroboration intensity. Heavy third-party mention work moves ChatGPT most and AI Overviews least. Conversely, a technical SEO project that lifts twenty pages into the top ten will move AI Overviews substantially and ChatGPT hardly at all.

The practical consequence is that a single AI visibility number averaged across engines hides the two most useful facts: which engine you are failing, and therefore which kind of work is missing.

Reading the pattern across engines

The combination of results tells you more than any single engine, and four patterns come up repeatedly.

Present in AI Overviews, absent from ChatGPT and Perplexity. Almost always rendering. Googlebot executes JavaScript and the others do not, so this pattern is close to a definitive diagnosis on its own. Fix rendering and the other two follow.

Present in Perplexity, absent from ChatGPT. Usually timing at first — Perplexity reflects changes in days and ChatGPT in weeks. If it persists past two months, the cause is corroboration: independent sources do not describe you consistently enough for ChatGPT to name you.

Named by ChatGPT, absent from AI Overviews. Your brand is well described by third parties while your own pages do not rank well enough to enter Google’s candidate set. A corroboration success sitting on an SEO weakness, and the fix is ordinary ranking work.

Absent everywhere. Check tier one before anything else. Universal absence with decent organic performance points to a mechanical fault affecting all crawlers — a CDN block, or content that only exists after JavaScript.

Budgeting effort across the three

A reasonable default split for a team with limited capacity: sixty percent on the shared foundation, twenty-five percent on corroboration aimed at ChatGPT, and fifteen percent on passage-level work aimed at AI Overviews.

The foundation dominates because it pays into all three simultaneously and because failing it makes the other two impossible. The ChatGPT allocation is second because it is the engine most buyers use and the one that responds slowest, so it needs the longest runway. AI Overviews get the least dedicated effort not because they matter least but because your existing SEO programme is already doing most of that work.

Adjust by where your buyers actually are. A developer-tools company will find Perplexity and ChatGPT disproportionately important; a consumer-facing business with heavy informational search will find AI Overviews are most of the exposure.

What not to conclude

Resist reading a single engine’s result as your overall position. Perplexity’s speed makes early programmes look successful and ChatGPT’s lag makes them look failed, and both readings are artefacts of measurement timing rather than findings about your visibility.

Measuring across three engines

Run the same fixed prompt set against all three, several times each, and report them separately.

Separate reporting is not a nicety. Because the engines respond at different speeds and to different inputs, a blended score moves for reasons nobody can attribute. Three lines on one chart — mention rate per engine over time — tells you what a single line cannot.

Store the raw answers in every case. Generated responses vary between identical runs, so a score without stored evidence cannot be audited or trended honestly. And record the cited URLs per engine, because the citation maps differ: the sources ChatGPT trusts in your category are frequently not the ones Perplexity uses, and that difference is where the corroboration brief comes from.

Our guide to tracking AI brand visibility covers the methodology in full.

Sources

• Vercel and MERJ — analysis of 500M+ GPTBot fetches finding zero JavaScript execution; GPTBot downloaded JS files ~11.5% of the time without running them; same behaviour for ClaudeBot and PerplexityBot.
• SERP feature tracking, 2026 — AI Overview presence averaging ~32.5% of monitored queries within a 31.02%–34.40% band.
• Google — 2.5 billion monthly active users for AI Overviews and 1 billion for AI Mode, stated 19 May 2026.

Frequently asked questions