How to Get Recommended by ChatGPT, Perplexity & Claude: The 2026 GEO Playbook

Why AI recommendations work differently from Google

Google ranks pages. AI engines synthesize an answer from a handful of sources they trust, then name a shortlist. Crucially, Google rank is a poor predictor of AI citations — a large share of ChatGPT-cited pages don't rank in Google's top results. And engines diverge: ChatGPT favours encyclopedic, comprehensive coverage, while Perplexity rewards recency and concrete examples. Optimising for one is not optimising for all.

Your job is to influence step two: be present, in the right words, on the sources the engine retrieves. Everything below is about earning your place there.

The seven-step playbook

1. Let the AI crawlers in

Check your robots.txt allows GPTBot, OAI-SearchBot, PerplexityBot, ClaudeBot, and Applebot. Blocking any one means you're invisible to that engine no matter how good your content is. This is the single most common own-goal.

2. Front-load a quotable answer

At the top of every important page, put a self-contained answer of 40–60 words that includes at least one specific statistic or named entity. That's the passage LLMs lift as a canonical citation. Bury the answer below marketing fluff and it won't get quoted.

3. Add schema — especially FAQPage

Structured data helps engines parse you as an entity. FAQ quality carries the highest individual signal weight. Use FAQPage schema, keep each answer 40–60 words, and match the question text to a visible on-page header character-for-character. Add Organization and Product schema so your entity is unambiguous.

4. Get onto the sources AI already cites

This is the highest-leverage move and the one most people skip. Identify the roundups, comparison pages, and directories the engines cite for your category, and get listed or reviewed there. Reddit matters more than you think — Perplexity, Gemini and Google AI Overviews weight relevant subreddit threads heavily, and a genuine long-term presence in 2–3 subreddits compounds.

5. Publish an llms.txt

Add a spec-compliant /llms.txt that points engines to your priority content. Honest caveat: llms.txt is a proposed convention with no standards-body backing, and no major provider has publicly confirmed reading it in production. But it's a low-cost, positive signal — 30 minutes of work — so do it.

6. Publish comparison pages for the exact buyer question

Create answer-first pages titled like the questions buyers ask ("best [category] for [use case]") and include plain, extractable comparison bullets that name you alongside rivals. These pages are what engines pull from when a buyer asks for a shortlist.

7. Measure monthly

Citations move slower than rankings. Pick 10–20 target prompts and run them monthly across ChatGPT, Perplexity, Gemini and AI Overviews, tracking your share of answers over time. (This is exactly what AIRanQ automates — a weekly scan across 16 engines with the fixes generated for you.)

Do these first (in order)

You can't do everything at once. If you only have a few hours this week, spend them in this order — highest leverage first:

1. 1

   Unblock the crawlers (30 min)

   Fix robots.txt so every AI crawler is allowed. This is the one that makes all your other work count.

2. 2

   Front-load answers on your top 5 pages (1 day)

   Add a 40–60 word extractable answer, with a specific number or named entity, to the top of your most important pages.

3. 3

   Add FAQPage schema (half a day)

   The single highest-weight structured-data signal. Match questions to visible headers exactly.

4. 4

   Get on 3–5 cited sources (ongoing)

   Directories, roundups, and 2–3 relevant subreddits. This is slow but compounds — and it's what actually moves citations.

Common mistakes that keep you invisible

• Blocking AI crawlers with a generic Disallow: / meant to stop scrapers — it stops the engines too.
• Burying the answer below hero copy, so there's no clean passage to quote.
• Optimising only for Google and assuming AI follows — it doesn't.
• Chasing one engine. ChatGPT and Perplexity reward different things; track and optimise for the set your buyers use.
• Expecting overnight results. Citations follow trust, which takes weeks. Measure monthly and stay consistent.