- • Why this check matters more than it sounds
- • Step 1: Write your real customer questions
- • Step 2: Ask each question in each tool
- • Step 3: Record what you find, precisely
- • Comparison: what each result actually means
- • 7 Detailed FAQs
You can check whether ChatGPT, Gemini and Perplexity recommend your business in about 45 minutes, for free, using nothing but the standard versions of those tools and a simple note-taking document. Ask the real questions your customers would ask, record whether your business is named, and you'll know exactly where you stand — most small business owners have never actually looked.
Why this check matters more than it sounds
Customers increasingly ask an AI assistant for a recommendation before they ever open Google. "Who's a good electrician near me," "which accountant should I use for a small business," "best solar installer in [town]" — these are now genuinely common AI prompts, not hypothetical ones.
When an AI names your business, two things happen at once: you land on the shortlist, and you borrow the model's own credibility. When it doesn't name you, neither happens — and unlike a weak Google ranking, there's no results page to scroll down and stumble onto you anyway. You're simply absent from the conversation.
Most small business owners have never checked. This walks through exactly how.
What you'll need
Nothing paid:
- Free access to ChatGPT, Perplexity and Gemini (and Google, to check for an AI Overview)
- A simple document or spreadsheet to record results
- A short list of your main local competitors
- 45 minutes, uninterrupted
Step 1: Write your real customer questions
List 10–15 questions a genuine customer would actually ask an AI assistant. Mix three types:
- Category + location: "best [your trade] near [your town]"
- Problem-based: "who can fix [specific problem] near me," "how do I find a reliable [your service]"
- Comparison/shortlist: "top [your trade] companies in [area]," "which [your service] has the best reviews"
Write them as full, natural questions — the way a person actually talks to an AI, not fragments. These prompts are your new keywords.
Step 2: Ask each question in each tool
Test every question in ChatGPT, Perplexity and Gemini, and check Google for an AI Overview. Use a fresh chat each time so earlier conversation doesn't bias the answer, and avoid a signed-in account the AI might already associate with your brand.
Read each answer properly. You're looking for three things: is your business there, are your competitors there, and is anything about your business wrong.
Step 3: Record what you find, precisely
Use four categories for each result:
- Cited — named with a clear source attribution or link. The strongest outcome.
- Mentioned — named, but without clear attribution.
- Absent — you didn't come up at all.
- Inaccurate — you appeared, but something was wrong (old address, wrong service, outdated details).
A page full of "Absent" results is exactly what this exercise exists to surface. Inaccurate results deserve extra attention — wrong information actively misleads a potential customer, which is often worse than not appearing at all.
Comparison: what each result actually means
| Result | What it means | What to do about it |
| Cited | You're the trusted, named answer | Keep content fresh, maintain the advantage |
| Mentioned | You're on the radar, not fully credited | Strengthen entity signals and third-party mentions |
| Absent | You're invisible for this question | Create clear content answering it directly |
| Inaccurate | You're actively misrepresented | Fix the source feeding the wrong information, urgently |
Step 4: Check your competitors, not just yourself
For every question where you're absent, note who did appear. Then look at why. Do they have a page that directly answers that exact question? Strong, recent reviews? Consistent listings across directories? Content that's clearly been updated recently?
This turns the exercise from a scoreboard into a map. Whoever's showing up is demonstrating, in practice, what "citable" looks like in your category and area.
Step 5: Work out why you're missing
Absence almost always traces back to one of a handful of fixable causes:
- No content answering the question. An AI can't cite an answer that was never published.
- The answer exists but is buried. Your page might cover the topic but not plainly or early enough to be extracted.
- Inconsistent business details. Your name, category or location aren't consistent across the sites where you're listed, so the model isn't confident what you actually are.
- Thin third-party presence. Few directories, reviews or genuine mentions reinforcing that you're real and relevant.
- Stale content. Pages that haven't been touched in a long time, which AI answers tend to deprioritise.
Step 6: Turn it into a short action list
Convert what you found into a ranked list:
- Fix inaccuracies first. Wrong information is actively costing you — correct it at the source, whether that's your own site or a directory listing.
- Create direct-answer content for the high-value questions where you're absent but competitors show up.
- Tighten your entity signals — identical name, category, location and contact details everywhere.
- Build genuine third-party presence — real directories, real reviews, real mentions.
- Refresh your key pages on a regular schedule going forward.
That list is, in effect, your own starting GEO plan — the audit just showed you where to point it.
How often to re-check
Quarterly is a sensible baseline for a small business. AI answers shift as models update and as competitors publish new content, so a once-a-year glance misses too much. Check again sooner if you've just pushed a content or review campaign and want to see whether the AI has caught up.
Forty-five minutes tells you something no traditional rank tracker can: whether the AI tools your customers increasingly ask are recommending you, or quietly pointing them somewhere else. If the results were mostly "Absent" and you'd rather have someone fix that than DIY it, that's exactly what we do.
Check you business is showing on ai