How to get recommended by ChatGPT, Perplexity and Google's AI answers

Why your business is (or isn't) being recommended by AI assistants

ChatGPT, Perplexity and Google's AI answers don't "know" your business. When someone asks for a recommendation, each assistant retrieves pages using its own crawler and quotes what it finds. That's the whole mechanism — and it means being crawlable, consistent and citable beats any trick.

The three assistants don't even look in the same places. ChatGPT's search leans on Bing's index plus its own crawler, OAI-SearchBot. Perplexity runs PerplexityBot and maintains its own index. Google's AI answers ground in Google Search. Same question, three different retrieval systems — which is why we test all three rather than treating "AI visibility" as one thing.

The consumer version of this question — "is Perplexity accurate?" — has flipped into the owner's version: "how do I get recommended?" This post answers the owner's version.

On 7 October we published a ChatGPT-specific checklist. This post is the wider version: all three assistants, plus the mechanism behind each step, so you're not following instructions blind. If you only sell in one town and want the short form, start with that one; if you want to know why each step works and how the assistants differ, read on here.

The thesis, plainly: assistants recommend businesses they can retrieve and quote, not the ones with the loudest marketing. This is plumbing and evidence, not secrets.

Step 1: Confirm the assistants' crawlers can actually reach your site

If the bot is blocked, nothing else on this list matters. This is the non-negotiable first check.

Open your robots.txt — it lives at yoursite.com/robots.txt — and look for blocks on OAI-SearchBot, PerplexityBot, Googlebot and ClaudeBot. In our experience, most of the blocks we find on client sites are accidental rather than deliberate. Cloudflare made AI-bot blocking the default for new domains in 2025, and security plugins and hosts ship it too. It's easy to switch on without noticing, usually during a security review or a hosting migration.

Here's the part people miss: being visible to Googlebot doesn't mean visible to the others. Every assistant has its own crawler and its own retrieval layer. A site that ranks fine in Google can be completely invisible to Perplexity.

Run our free AI readiness check and it'll tell you in under a minute whether the AI crawlers can reach your site and what, if anything, is blocking them. No sign-up, and it's the same first test we run on every client site.

Step 2: Be findable where each assistant actually looks

Crawl access gets the bot to your door. Index coverage gets you onto the shelf it pulls answers from. This is the step most advice skips entirely.

ChatGPT leans on Bing's index. So: get indexed in Bing, and claim your Bing Places listing. In our experience most small businesses have never opened Bing Places — if you've only ever verified your Google Business Profile, this is likely where your ChatGPT visibility is thinnest. Check rather than assume.

Google's AI answers ground in Google Search. That means the standard plumbing: proper indexing, Search Console, and a complete Google Business Profile. If you rank in Google, you're already in the pool its AI answers draw from — though being in the pool isn't the same as being picked. If you're not sure whether Google has even indexed your service pages, start with how Google finds your pages and our Business Profile checklist.

Perplexity maintains its own index, and there's no equivalent of "submitting" to it. That makes two things matter more: crawl access (step 1) and third-party mentions (step 4), because Perplexity builds its picture of you from what it can retrieve, wherever that lives.

One honest caveat: this plumbing changes. We re-test our assumptions quarterly, and you should re-check yours too — if ChatGPT stops leaning on Bing, the Bing advice above changes with it.

Step 3: Keep your entity consistent everywhere

Assistants reconcile facts across your website, your Business Profile and the directories. When those sources disagree, you become unreliable to quote. Contradiction is the silent killer.

Keep one source of truth for your name, address, phone number, opening hours and service area — identical on your site, Google Business Profile, Bing Places and the directories that matter in your trade.

The most common failure we see on WordPress sites: a theme, an SEO plugin and a dedicated schema plugin all emitting LocalBusiness markup with conflicting facts. That's worse than one clean set, because now the assistant has to guess which version is true. The fifteen-minute fix is in our structured data plugin guide: pick one source of truth for your facts, disable or configure the other markup emitters, and re-test after every theme or plugin update, because schema breaks silently.

Two honesty notes on markup:

• Structured data helps assistants read who you are, but no tag can force a citation. As of October 2026 we have no documented evidence that assistants weight markup in their recommendations. Treat it as machine-readability hygiene, not a lever — the same verdict we gave in our structured data post.
• Don't pay anyone for an llms.txt file. There's no solid evidence any assistant uses it. We don't charge clients for publishing one, and neither should anyone else.

If your baseline indexing and titles need attention first, run our free SEO audit — no point optimising for AI answers if Google hasn't indexed your service pages.

Step 4: Earn mentions on the third-party sites assistants quote

Here's the uncomfortable truth for "recommend a plumber near me" style questions: the evidence layer isn't your website. It's reviews, directories and trade sites — the pages assistants retrieve and quote.

Know which sources are retrievable in your trade:

• Trades: Checkatrade, Which? Trusted Traders, and your trade body's directory
• Hospitality: TripAdvisor, OpenTable, Google Business Profile reviews
• Everyone: Yell and the sector directories your customers actually browse

Reviews are the relevance evidence. A customer writing "called them out as an emergency plumber in Leeds on a Sunday" tells every retrieval system — Google's and the assistants' — what you are and where you are. Influence it by being memorable; never script it.

Steady velocity beats bursts. We've watched steady profiles hold positions while burst profiles drift, across client accounts — and a trickle of recent reviews is far less likely to trip review platforms' spam filters than a clump of twenty after two years of silence. Don't buy or incentivise reviews: incentivised reviews breach platform rules and UK consumer law, platforms remove them, and a stripped profile is worse than a small one. We set out the evidence in our review ranking post.

Your own site still matters, but for a specific reason: assistants quote pages that state facts plainly. "We service boilers across LS6 and LS7, same-day callouts, Gas Safe registered" is quotable. "Delivering excellence in heating solutions" is not. Write service pages that a machine could read aloud as an answer.

This is the core of the AI search visibility work we run for clients — less cleverness, more evidence in the right places.

Step 5: Test monthly with real buying questions

This is the habit that separates a one-off checklist from an actual system.

1. Ask each assistant the questions real customers ask — "find a plumber in Harrogate", "best family dentist in Reading" — using your trade and town.
2. Log who gets cited, and from which source. A spreadsheet with five columns does it.
3. Compare across ChatGPT, Perplexity and Google's AI answers. Being known by one doesn't mean being known by all — that's the point of testing all three.
4. When a competitor is cited instead, find the source page. Then decide: earn a mention there, or beat it with better evidence elsewhere.
5. Re-test after any site, plugin or hosting change. Blocks reappear silently — a plugin update can undo step 1 without telling you.

Monthly is enough. Weekly is noise, because retrieval results shift and you'll chase ghosts.

What nobody can guarantee — and how to spot the sales pitch

Let's be plain about the limits, because the content currently ranking for this query mostly won't.

There is no paid placement in ChatGPT recommendations. Any offer of guaranteed placement is misleading. The same logic applies to Perplexity and Google's AI answers.

"Am I in the training data?" is unanswerable and mostly irrelevant. "Can each crawler reach and read my site today?" is checkable — and it's the one that matters.

Nobody outside the assistants' owners knows the ranking mechanics. Us included. What we can do is make you easy to find, easy to read and easy to trust, then measure what changes. Anyone promising more than that is guessing and charging you for the guess.

Watch for these four when the pitches arrive:

• llms.txt packages — no evidence assistants use it
• "AI SEO secret" courses — there are no secrets, only plumbing and evidence
• Guaranteed citations — nobody can guarantee what they don't control
• Schema packages sold as a citation switch — markup helps machines read you; it doesn't switch recommendations on

Your fifteen-minute start this week

Five actions, in order:

1. Run the free AI readiness check — see whether the crawlers can reach you
2. Check robots.txt for anything blocking OAI-SearchBot, PerplexityBot, Googlebot or ClaudeBot
3. Claim your Bing Places listing — if you haven't already
4. Fix any conflicting facts across your site, Business Profile and directories — one source of truth
5. Run one real buying question on each assistant and log the result

On timing: there's never a bad week to start, and the reason to start now is simply that the evidence layer takes months to build. Crawl access is fixed in an afternoon; a profile of steady reviews and consistent facts is not. The earlier it's in place, the longer it has to work.

If you'd rather we ran it, that's exactly what our AI search visibility service covers — and the first call is free: book it here.

Questions people ask

Found this useful? Share on LinkedIn · Email it to someone