Closing AI Visibility Gaps Against Your Competitors
An AI visibility gap is what you see when someone asks ChatGPT, Perplexity, Gemini or Google's AI Overviews a question in your field and the answer names a competitor but not you. It matters because more people now ask an assistant for a shortlist before they ever visit a website.
Assistants do not pick names at random: they lean on pages they can retrieve and on what other sites say about each business. This guide shows how to measure the gap fairly, trace the sources behind each rival mention, and work on the things that make your business easier to find, trust and quote.
How assistants decide which businesses to name
The Competitor Analysis includes visibility in AI answers alongside keywords, links and schema, because rivals increasingly compete for a place in a generated answer as well as a place on the results page. The mechanics differ by product, but most answers that name businesses draw on two things.
- Live retrieval. ChatGPT search, Perplexity, Gemini and Google's AI Overviews can fetch web pages at the time of the question and cite them. If your page is not retrievable or does not answer the question clearly, it cannot be cited.
- What the web says about you. Models learn from large amounts of text, and live answers often lean on round-ups, directories, reviews, forum threads and news. A business mentioned often, in the right context, is more likely to be named.
So a rival can be named because their own page answers the question well, because third-party pages keep listing them, or both. The fix depends on which, and the only way to find out is to look at the sources.
Building a fair prompt set
A prompt set is a fixed list of questions you ask every assistant, every time you measure. Keeping it fixed is what makes the numbers comparable month to month.
- Start from real questions. Use enquiry emails, phone notes, sales calls and Search Console queries. People ask assistants longer, more conversational questions than they type into a search box.
- Mix the types. Include recommendation prompts ("best boiler engineer in Chesterfield"), comparison prompts ("Brookside Heating or Rival A for a combi boiler"), and problem prompts ("my boiler is showing an F28 error, what should I do").
- Keep it manageable. Fifteen to thirty prompts is enough for a small business. More than that and you will stop running it.
- Record the conditions. Note the date, the assistant, whether you were signed in, and whether web search was on. Use a signed-out or fresh session where possible so your own history does not colour the answers.
| Prompt | Assistant | Run | Named | Cited sources |
|---|---|---|---|---|
| Best boiler engineer in Chesterfield | Perplexity | 1 | Rival A, Rival B | local-trades-roundup.example, rival-a.example |
| Best boiler engineer in Chesterfield | Perplexity | 2 | Rival A, you | local-trades-roundup.example, review-site.example |
| Best boiler engineer in Chesterfield | AI Overview | 1 | Rival B | rival-b.example/boiler-repair, directory.example |
After a few rounds you can count, for each prompt and assistant, how often each business was named. That share of mentions is your measure of the gap.
Tracing the sources behind a rival's mentions
The sources column is the most useful part of the log. Perplexity lists its sources for every answer, Google's AI Overviews show linked sources, and ChatGPT shows citations when it searches the web. Group the cited pages into types:
| Source type | What it suggests | Response |
|---|---|---|
| The rival's own page | Their page answers the question more clearly than yours | Improve your matching page; see the section on quotable pages |
| "Best of" round-ups and local lists | Third-party lists drive the recommendation | Find out how the list is compiled and whether you can be considered |
| Review platforms and directories | Review volume and consistency matter | Complete your profiles and ask real customers for reviews |
| Forums and community threads | People recommend the rival in discussions | Take part helpfully where you have expertise; do not astroturf |
| News and trade press | The rival is covered as an authority | Offer genuine stories and expert comment |
Most gaps turn out to be a mix, but one type usually dominates. That is where to start.
Ruling out access problems
Before working on mentions, make sure the assistants can actually read your site. Check robots.txt
for rules affecting the crawlers that fetch pages for AI answers.
User-agent: OAI-SearchBot
Allow: /
User-agent: PerplexityBot
Allow: /
User-agent: Googlebot
Allow: /- OpenAI uses
OAI-SearchBotfor ChatGPT search results andGPTBotfor collecting training data. Blocking GPTBot does not block OAI-SearchBot; they are controlled separately. - Perplexity uses
PerplexityBotto index pages for its answers. - Google's AI Overviews use pages crawled by Googlebot for Search. The
Google-Extendedtoken controls some uses in Gemini models but does not remove pages from Search or AI Overviews.
Also check your firewall or CDN. Bot protection settings sometimes block AI crawlers even when robots.txt
allows them. And make sure your key content is in the HTML the server returns, not only rendered later
by JavaScript, since not every crawler runs scripts. For more, see fixing
AI crawler access.
Making your pages easy to quote
When the rival's own page is the cited source, compare it with yours for the same question. Pages that get quoted tend to share a few habits.
- The answer comes first. A direct one- or two-sentence answer near the top, then the detail. Assistants often lift that opening.
- Specific facts in plain text. Prices, areas covered, response times, accreditations and opening hours written out, not buried in images or PDFs.
- Clear headings that match questions. "How much does a boiler service cost in Chesterfield?" is easier to match than "Our pricing".
- Tables for comparisons. Structured comparisons are easy to extract accurately.
- Evidence of who you are. Named people, qualifications and real examples of your work.
The content extractability guide goes deeper on page structure. The aim is not to write for machines; pages that are clear for a hurried reader are also clear for an assistant.
The fix, step by step
Step 1
Write a fixed set of real customer prompts
Collect fifteen to thirty questions from enquiry emails, sales calls and Search Console queries, phrased the way people talk to an assistant. Include recommendation, comparison and problem-solving prompts, and name your area where location matters. Save the list and use exactly the same wording every time you measure, because changing prompts between rounds makes it impossible to tell whether the gap with your competitors is really closing.
Step 2
Run each prompt several times and log who is named
Ask each prompt in ChatGPT with web search, Perplexity, Gemini and Google Search where an AI Overview appears, at least three times each, using a signed-out or fresh session where possible. Record the date, the assistant, the run number and every business named. Answers vary from run to run, so the share of runs in which each competitor appears is a much fairer measure than any single answer or screenshot.
Step 3
Collect the sources behind every rival mention
For each answer that names a competitor, copy the cited sources into your log. Group them as the rival's own pages, round-ups and local lists, review platforms and directories, forums, or news and trade press. Count how often each group appears. The dominant group tells you whether the rival is winning through their own content, through what others publish about them, or through both, and so where your effort should go first.
Step 4
Rule out access problems on your own site
Check robots.txt for rules that block OAI-SearchBot, PerplexityBot
or Googlebot, and check your firewall or CDN bot settings, which can block crawlers even
when robots.txt allows them. Confirm the key facts on your main pages appear in the HTML the server sends
rather than only after JavaScript runs. A site that assistants cannot fetch will stay invisible whatever
else you improve, so clear this before anything else.
Step 5
Earn a place in the sources assistants already use
Work through the third-party sources that keep citing rivals. Find out how each round-up or local list is compiled and whether you can be considered, complete and correct your directory and review profiles, ask genuine customers for reviews on the platforms that appear most, and offer trade press a real story or expert comment. Never pay for undisclosed placements or post fake reviews; the aim is to be mentioned honestly where people already look.
Step 6
Make your own pages easy to quote
Where a rival's own page is the cited source, compare it with your matching page. Put a direct answer in the opening lines, use headings that mirror the questions people ask, write prices, areas, hours and qualifications as plain text, and use tables for comparisons. Add named people and real examples of your work. Clear pages help hurried human readers as much as assistants, so none of this work is wasted if AI tools change.
Step 7
Make your business easy to identify
Use
exactly the same business name, address, phone number and description on your site, your Google Business
Profile, directories and social profiles. Add Organization or LocalBusiness structured data with a sameAs
list pointing to your official profiles. Consistent details help assistants recognise that mentions across
different sites refer to the same business, which reduces the chance that a rival with a clearer identity
is named in your place.
Step 8
Re-run the same prompt set on a schedule
Repeat the full prompt set monthly or quarterly with the same assistants, the same number of runs and the same session conditions. Compare each competitor's share of mentions and the sources cited against earlier rounds. Changes in third-party sources and model updates take time to show, so look for a trend over several rounds rather than reacting to one month, and re-run the Competitor Analysis alongside for the wider picture.
Measuring progress without fooling yourself
- Share of mentions per prompt. The percentage of runs in which you are named, next to each rival's percentage.
- Sources that now include you. A new mention in a round-up or directory that assistants often cite is a leading indicator, even before your share moves.
- Accuracy of what is said. Being named with the wrong prices or an old address is a problem of its own; see fixing AI hallucinations.
- Referral traffic. Visits from chatgpt.com, perplexity.ai and similar referrers appear in most analytics tools, although some AI traffic arrives without a referrer.
1. Why does ChatGPT recommend my competitor and not me?
Usually because the sources it draws on mention the competitor more often or more clearly. When ChatGPT searches the web, look at the citations: if they are round-ups, directories or review sites listing the rival, the gap is about third-party mentions. If the rival's own page is cited, their page probably answers the question more directly than yours. Also check that your site does not block OAI-SearchBot, since a page it cannot fetch cannot be cited.
2. How do I measure AI visibility against competitors fairly?
Use a fixed list of real customer questions, ask each one in the same assistants several times with the same settings, and record every business named and every source cited. Work out the share of runs in which each business appears. Answers vary between runs, users and locations, so a single test or screenshot is not reliable. Repeat the same process monthly or quarterly so the numbers can be compared over time.
3. Does blocking GPTBot stop my site appearing in ChatGPT?
Not necessarily. OpenAI uses separate crawlers: GPTBot collects content that may be used for training, while OAI-SearchBot fetches pages for ChatGPT search results. They are controlled by separate robots.txt rules, so you can block one and allow the other. If you want to be cited in ChatGPT search answers, make sure OAI-SearchBot is allowed and that your firewall or CDN is not blocking it regardless of what robots.txt says. Test by checking your server logs for its visits.
4. Will adding schema markup get me into AI answers?
It can help assistants and search engines identify your business and understand your pages, particularly Organization or LocalBusiness details with consistent sameAs links to your official profiles. It is not a switch that earns citations on its own. What gets businesses named is mostly clear, specific content on their own pages and honest mentions on sites the assistants already use as sources. Treat structured data as a supporting signal kept accurate and consistent.
5. Can I pay to be included in AI recommendations?
There is no general way to pay the main assistants for organic recommendations. Paying for undisclosed placement in the round-ups and lists that assistants cite is risky: it can breach advertising rules and platform policies, and in the UK fake or incentivised reviews that are not disclosed fall under consumer protection law. This is general information, not legal advice. Earn mentions through genuine reviews, real coverage and helpful content instead.
6. How long does it take to close an AI visibility gap?
Expect months. New mentions on third-party sites have to be published and then retrieved or learned by the assistants, and model updates arrive on their own schedules. Improving your own pages can show sooner in assistants that search the web live, once the pages are recrawled. Track your share of mentions across several rounds of the same prompt set rather than expecting a change after one round of work, and keep the conditions identical each time.
7. Are AI Overviews controlled by the same rules as normal Google results?
AI Overviews appear in Google Search and draw on pages Google has crawled and indexed for Search, so the same basics apply: your pages must be crawlable by Googlebot, indexable and helpful. The Google-Extended robots.txt token affects some uses in Gemini models but does not remove your pages from Search or AI Overviews. To appear as a linked source, your page needs to answer the question clearly enough to support the summary.
Our Competitor Analysis compares your site with named rivals across keywords, content, backlinks, technical health, schema and visibility in AI answers, so you can see where competitors are named and you are not, and decide which gaps to close first.
Open Competitor Analysis Fixes →