Guide

How to check what ChatGPT says about your business (and fix it)

Published 16 July 2026 · AI Advertising

TL;DR: Ask ChatGPT the questions your customers ask, three ways: category ("best electrician in Penrith"), brand ("is [your business] any good?"), and comparison ("[you] vs [competitor]"). You'll land in one of four outcomes: recommended, mentioned, misdescribed, or invisible. Each has a specific fix, and most of the fixes are about making your business easier for machines to read and corroborate.

The ten-minute self-test

Open ChatGPT (the free tier is fine, but note it answers from an older snapshot of the web unless it searches). Ask these, substituting your details, and save the answers:

The category question: "Who's the best [your trade/service] in [your suburb or region]?" Then again for the neighbouring suburb, and again phrased as "who should I use for…".
The brand question: "Is [your business name] any good?" and "What do you know about [your business name] in [your area]?"
The comparison question: "[Your business] vs [your main competitor], which should I choose?"

If you can, repeat the category question in a fresh chat once or twice. AI answers vary between sessions, and what you're looking for is the pattern, not a single response. Do the same in Google's AI Mode and Perplexity if you have another ten minutes: they retrieve from the live web, so they show you the "searching" version of the problem, while ChatGPT without browsing shows you the "model memory" version. You want to be visible in both.

The four outcomes, and what each one means

1. You're recommended

Congratulations, you're the incumbent. Your job now is defence: keep your reviews, facts and content consistent, because the AI's confidence in you is built on corroboration, and corroboration decays when details drift. Re-run the test monthly. Incumbents who coast get replaced quietly.

2. You're mentioned, but not first

The AI knows you exist but prefers someone else. Look closely at how it describes the winner: it's usually specific ("known for same-day service", "strong reviews for strata work"). Those phrases come from somewhere: review text, site copy, directory listings. Your competitor's public footprint answers the question better than yours does. The fix is to close the specificity gap: publish clear service pages that say exactly what you do, for whom, where, and back them with review volume on the platforms the AI cited.

3. You're described wrong

Wrong services, wrong suburb, a confusion with a similarly named business. This is an entity problem: the web's facts about you disagree, so the model blends or guesses. The fix is consistency work: one canonical description of your name, services and location, repeated exactly on your site's structured data, your Google Business Profile, directories and socials. Machines resolve contradictions in favour of whoever is most consistent, so be that business.

4. You're invisible

The AI has never heard of you, or answers only with national chains. Almost always this pairs with technical findings: AI crawlers blocked in robots.txt, no structured data, content that only exists inside JavaScript, no llms.txt, thin pages that never answer a question directly. Start with the technical foundations (they're quick), then build the answer-shaped content that gives the AI something worth citing.

The fixes, in the order we'd do them

Unblock the crawlers. Check robots.txt for GPTBot, ClaudeBot, PerplexityBot, Google-Extended. Check your CDN or host isn't blocking "AI bots" on your behalf. This one setting can be the whole problem.
Publish your facts in machine form. JSON-LD structured data on every key page: organisation, services, area served, FAQs. Add an llms.txt. Give the machines a version of you they can quote without guessing.
Answer real questions on real pages. One page per question your customers actually ask, question in the heading, answer in the first sentence. This is what retrieval-based AI cites.
Fix the disagreements. Same name, same services, same suburbs, everywhere. Update the stale directory listings you forgot you had.
Build the review base. AI assistants lean hard on review signal, volume, recency and specifics. Ask every happy customer, and make it easy.
Re-test monthly and keep the transcripts. The test at the top of this page is your scoreboard. Improvement shows up there before it shows up in revenue, which is exactly why it's worth tracking.

A note on "guaranteed ChatGPT rankings"

You'll meet vendors selling guaranteed placement in AI answers. There is no such product. Nobody outside OpenAI controls ChatGPT's answers, and the honest version of this work is influence, not control: make yourself the easiest, safest, best-corroborated answer, then verify with your own transcripts. Anyone promising more than that is charging you for a coin flip.

Want it done for you?

Our free AEO audit runs this exact test plus the technical checklist against your site, and sends you the transcripts, your pass/fail on each signal, and the three fixes that matter most for your situation, within one business day.