Short answer: In most cases the AI simply has more to go on with them. Their name crops up on other people’s websites and their pages say clearly what they sell. Often their site is easier for AI crawlers to get into than yours, too. So if a competitor keeps popping up in ChatGPT or a Google AI Overview while you don’t, that’s usually where the gap lies. Nothing here guarantees a citation. Most gaps can be traced, though.
We hear some version of this most weeks. Someone asks ChatGPT for the best kitchen fitter near Newry and a competitor comes up instead. Then Google shows the same firm in its AI Overview. It stings.
It’s easy to assume they’ve found a clever trick. Honestly, that’s rarely it. Below are the usual suspects, roughly in the order we’d check them. If AI search is new to you, our complete guide to AI search visibility is worth reading alongside this.
Check What’s Really Happening First
Get a proper picture first. One screenshot from one chat proves little.
ChatGPT
Ask ChatGPT the same question on Monday and again on Thursday and you may well get a different list. We run each prompt three or four times, in a fresh chat each go, and only pay attention to names that keep reappearing. One stray mention? Ignore it.
Google AI Overviews and AI Mode
Google describes both as rooted in its core ranking systems. The catch is they can run on different models, so a firm linked in one might be missing from the other. A private browser window gives the cleanest view. Their traffic sits within the normal Web figures in Search Console.
Copilot, Perplexity and Gemini
Since February 2026 Bing Webmaster Tools has had an AI Performance report (still a public preview) showing how often Copilot and Bing’s AI answers cite your pages. Perplexity uses its own crawler, PerplexityBot. Gemini is worth testing too.
Keep evidence and hunches apart
Log the prompt, the platform and whoever got named. That’s your evidence. Any reason you come up with for a competitor being picked stays a hunch until you’ve tested it below.
Eleven Reasons a Competitor Shows Up and You Don’t
There’s no public weighting for any of these. It’s just the order we tend to use.
1. Other websites talk about them
Possible cause: A fair amount of what AI answers say about a business seems to come from other people’s websites. A rival who’s been written up in the local paper or a couple of trade directories simply has more for the AI to draw on.
How to check: Search their brand name in quotes with -site: and their domain added, so their own pages drop out. Then do yours. Crude, but usually revealing.
What to fix: Start with the directories your customers actually use. Then offer a genuine local story to a paper or trade magazine, and ask suppliers if they’ll mention you.
2. Respected sources mention them
Possible cause: Our working assumption is that a mention from a recognised industry body counts for more than one on a thin, auto-generated directory. We’ve never seen an AI company confirm it, mind.
How to check: The clue is usually sitting in the answer itself. Click whatever sources the AI linked beside your competitor’s name and you’ll see where it picked them up.
What to fix: Joining a relevant professional association is often the quickest win.
3. Their pages actually answer the question
Possible cause: Google’s guidance for its AI features keeps coming back to unique, satisfying content. Say a competitor’s page gives a rough price for a fitted kitchen in Newry and yours just says “get in touch”. Guess which gets quoted.
How to check: Take each prompt where they appeared and hunt for the page on your own site that answers it within the first couple of lines. Often there isn’t one.
What to fix: Put the answer up top and the caveats after it.
4. They cover more of the topic
Possible cause: Google says AI Overviews and AI Mode may run several related searches across subtopics. A site covering the obvious follow-up questions has more ways in.
How to check: We usually jot down the six or so things customers ask before they buy, then go through both websites ticking off which get answered.
What to fix: Any question you couldn’t tick is a page waiting to be written. Link it from the related service page so people can find it.
5. Their details are consistent
Possible cause: An AI has to work out that “Smith Joinery” in one directory and “Smith & Sons Joinery Ltd” in another are the same firm. An old address still floating around makes it harder again.
How to check: Type your business name into Google and scroll page one slowly. You’re hunting for an old address, a former trading name or a dead phone number.
What to fix: Pick one way of writing your details and stick to it on every listing. Your Google Business Profile is the obvious first stop, as Google itself recommends keeping it current.
6. Structured data describes them clearly
Possible cause: Schema markup helps search engines read who you are. Google says no special schema is needed for its AI features, though, so it won’t move things on its own.
How to check: Drop your homepage address into validator.schema.org. It takes about a minute.
What to fix: If nothing useful shows up, ask whoever built your site to add Organization or LocalBusiness markup that matches what visitors actually see.
7. AI crawlers can get into their site
Possible cause: This one catches people out more than anything else here. A security plugin, host firewall or broad robots.txt rule can block OAI-SearchBot, PerplexityBot or Bingbot unnoticed. OpenAI says sites blocking OAI-SearchBot won’t be shown in ChatGPT search answers.
How to check: Add /robots.txt to the end of your web address and have a look. A “Disallow: /” line under “User-agent: *” is the classic culprit. Host-level bot protection won’t show up there, so it’s worth a quick email to your hosting company too.
What to fix: Allow the search crawlers you want to appear in. GPTBot, which OpenAI uses for training, can still be blocked separately.
8. Their key pages are indexed
Possible cause: Google won’t use a page as a supporting link in its AI features unless it’s indexed and eligible to show a snippet.
How to check: Search Console’s Pages report flags whatever Google has left out, and URL Inspection checks one page at a time. Bing Webmaster Tools has its own version, which people often forget about.
What to fix: Usually it’s a stray noindex tag or a canonical pointing the wrong way. Our SEO tip on finding pages Google isn’t indexing covers the fix.
9. They publish something nobody else has
Possible cause: Plenty of service pages read as if they were written for the whole industry. Why would an AI pick yours over a dozen that say the same thing?
How to check: Pull up your five most important pages and be a bit ruthless. Is there one fact, figure or opinion on them a reader couldn’t find elsewhere?
What to fix: Add things only you know. A south Belfast physio clinic could publish typical recovery timelines, with frank caveats. That’s hard to copy.
10. Their reputation is easy to see
Possible cause: Reviews probably play some part in whether a platform treats a business as worth recommending, though nobody outside these companies knows how much.
How to check: Compare your Google reviews with theirs, looking at recency and replies as well as totals.
What to fix: Ask for a review after every job and reply to each one. Yes, even the grumpy ones.
11. They say what they do in plain words
Possible cause: A tagline like “solutions for growth” gives an AI almost nothing to match against a request such as “emergency boiler repair in Lisburn”.
How to check: Read your homepage out loud to someone who doesn’t know the business. Afterwards, can they tell you what you sell and where you work?
What to fix: Name each service in customers’ words, mention the towns you cover and give each one its own page.
Cause and Fix at a Glance
| Possible cause | Quick check | First fix |
|---|---|---|
| Few third-party mentions | Brand search minus own domain | Directories, press, partners |
| No respected sources | Sources the AI links to | Associations, awards, comment |
| No direct answers | Match prompts to pages | Answer first |
| Thin topic coverage | Map buyer questions | Focused, linked pages |
| Inconsistent details | Compare listings | One version everywhere |
| Missing schema | Schema validator | Organization or LocalBusiness |
| Crawlers blocked | robots.txt, host settings | Allow search crawlers |
| Pages not indexed | Search Console, Bing | Fix noindex, canonicals |
| Nothing original | Honest uniqueness test | First-hand detail |
| Weak reputation | Compare reviews | Review habit |
| Vague descriptions | Read-aloud test | Plain service pages |
Quick Self-Assessment Checklist
Run down this list and tick whatever’s already sorted.
- robots.txt and the firewall let AI search crawlers through
- Key service pages are indexed in Google and Bing
- Each core service has a page of its own, written in plain English
- Your name, address and phone number match across the main listings
- Independent websites mention you
- The questions customers ask most get answered somewhere on the site
- Recent reviews, with replies
- You’ve logged five prompts on three AI platforms
Ticked fewer than five? The plan below is where we’d begin.
Where to Start When Time or Budget Is Tight
Fix access first. There’s no point polishing a page an AI crawler can’t read.
This week
Spend an afternoon on the basics. Check robots.txt and your host’s bot settings, confirm your main pages are indexed in Google and Bing, then run your baseline prompts.
This month
Tidy up your business details everywhere, then rework your three most important service pages so each opens by saying plainly what you do.
Over the next two to three months
This is the slow bit. Chase genuine mentions, add original detail and keep the reviews coming. Re-run your prompts monthly to see what’s moved.
What’s Confirmed and What’s Still a Hunch
A few points here come straight from platform documentation, mainly Google’s indexing requirement (no special schema or AI text files needed) and OpenAI’s crawler rule. Bing’s Copilot citation report is first-party data too.
The rest comes from what we and others in the industry have observed. No AI company has published it as a formula, though, and we’d be wary of anyone who says otherwise.
When It’s Worth Getting Help
Plenty of this is DIY. It gets fiddly across several competitors and platforms, or when a crawler problem is buried in your hosting. Our AI Search Optimisation service handles that analysis, the technical checks and the slower authority work.
Book a free call and we’ll compare you with a couple of competitors and talk you through the gaps. We won’t promise citations, because nobody honestly can.
Frequently Asked Questions
How can I check whether a competitor appears in AI search?
Pick five to ten prompts a real customer might type and try each one in ChatGPT, Google, Copilot and Perplexity. Use a fresh chat or private window every time. Write down who gets named or linked, then repeat the whole thing a few days later.
Why does ChatGPT recommend certain businesses?
ChatGPT can only cite pages its search crawler is allowed to reach, and it tends to lean towards sources that answer the question plainly. Beyond that, OpenAI hasn’t published its full list of factors. Anyone who claims to know them all is guessing.
How long do AI visibility improvements take?
Some fixes register fast. OpenAI has said robots.txt changes take around 24 hours to be picked up for search, for instance. Content and reputation work is slower, and we’d usually judge it over three to six months. Even then there’s no guarantee.
