If you have ever asked who does ChatGPT recommend in my industry, the useful question is not whether your company appears once.
It is whether the same competitors keep appearing when you ask the kinds of questions a real buyer would ask.
That is the Ranked Not Recommended problem in its simplest form.
Your company may rank well in Google.
Your website may be technically healthy.
Your brand may already be known in the sector.
But when a buyer asks an AI assistant:
Which companies should I investigate for this requirement?
three other names may repeatedly appear before yours.
You can test that yourself.
You do not need a complicated dashboard.
You do not need fifty prompts.
Start with six buyer-style questions across four AI assistants and record what happens.
The test will not give you an official AI ranking position. No major assistant publishes a universal business-recommendation league table.
What it gives you is more useful:
a repeatable snapshot of which companies keep getting named for the commercial questions you care about.
Table of Contents
- What this test is actually measuring
- Use buyer prompts, not brand prompts
- The six prompts to run
- Run the same prompts across four assistants
- How to record the results
- What does a repeat competitor name mean?
- Pattern 1: You are absent entirely
- Pattern 2: You appear occasionally
- Pattern 3: You appear with outdated or incorrect details
- Look at the evidence, not just the names
- What should you do with the result?
- Frequently Asked Questions About AI Competitor Visibility
- Run the test yourself
What this test is actually measuring
Do not treat this as a scientific market-share study.
You are testing something narrower:
When an assistant receives a commercially relevant question, which businesses does it choose to mention?
That matters because AI assistants can use web information and external sources when answering current questions.
ChatGPT Search can search the web and provide linked or cited sources in its answers.
Claude’s web-search capability similarly retrieves current web information and returns source citations.
Perplexity describes its search experience as synthesising information from multiple web sources and linking users back to the original material.
Perplexity: How Pro Search works
That makes the sources behind the recommendation almost as interesting as the recommendation itself.
Use buyer prompts, not brand prompts
Do not start with:
Is Lobo Digital a good agency?
Or:
Tell me about ABC Facilities Management.
You have already supplied the company name.
That tests brand retrieval.
It does not test discovery.
Instead, imagine the buyer has never heard of you.
What would they actually type?
A buyer might ask:
Which agencies specialise in GEO and AEO for UK B2B companies?
Or:
Which commercial cleaning companies cover multi-site estates in London?
Or:
Which cybersecurity consultancies work with mid-market UK businesses?
The question should describe the requirement, not the supplier.
That is the core rule.
The six prompts to run
Replace the brackets with your own market, service and customer type.
Broad Category
Which [type of company] providers should a UK business consider?
Commercial Requirement
Which companies can help a [buyer type] with [specific problem or service]?
Sector Relevance
Which [service] providers have experience working with [industry or sector]?
Geography
Which [type of company] providers operate across [location or region]?
Evidence or Qualification
Which [type of company] providers have evidence of experience with [specific requirement, accreditation or contract type]?
Shortlist
Give me a shortlist of [type of company] providers suitable for [specific buyer requirement].
Do not optimise the wording until your own company appears.
That defeats the purpose.
Use the language a buyer is likely to use.
Run the same prompts across four assistants
For a useful baseline, use the same six prompts across:
- ChatGPT
- Gemini
- Claude
- Perplexity
That gives you:
6 prompts × 4 assistants = 24 answers
Do not change the wording between platforms.
You are looking for patterns.
Use a fresh conversation for each test where practical.
Also record the date.
AI answers can change as:
- web sources change
- indexes refresh
- models change
- retrieval systems change
- location or context changes
- personalisation settings differ
Gemini, for example, now documents personalisation features that can use previous Gemini chats, connected Google content and user preferences in eligible experiences.
Google: Gemini personalisation
This is one reason the aim is repeatability, not pretending one result is permanent.
How to record the results
Keep it simple.
You need columns for:
| Prompt |  ChatGPT |  Gemini |  Claude |  Perplexity |
|---|---|---|---|---|
| Prompt 1 | Companies named | Companies named | Companies named | Companies named |
| Prompt 2 | Companies named | Companies named | Companies named | Companies named |
| Prompt 3 | Companies named | Companies named | Companies named | Companies named |
| Prompt 4 | Companies named | Companies named | Companies named | Companies named |
| Prompt 5 | Companies named | Companies named | Companies named | Companies named |
| Prompt 6 | Companies named | Companies named | Companies named | Companies named |
Then create a simple tally.
For example:
Competitor A: 16 mentions
Competitor B: 11 mentions
Competitor C: 8 mentions
Your business: 2 mentions
Do not turn those numbers into a formal “AI ranking score”.
They are simply the result of your defined test set on that date.
But the pattern can still be commercially useful.
What does a repeat competitor name mean?
Suppose one competitor appears in:
- five of six ChatGPT prompts
- four Gemini prompts
- four Claude prompts
- all six Perplexity prompts
Do not conclude:
They are the number-one AI-ranked company in the sector.
You do not have evidence for that.
The defensible conclusion is:
Within this test, that company is being associated with the category more consistently than we are.
Now investigate why.
Look at:
- its service pages
- sector pages
- case studies
- third-party coverage
- reviews where relevant
- company information
- structured content
- accreditations
- authoritative listings
- citations appearing in the AI answers
The tally tells you where to investigate.
It does not tell you the cause.
For the wider framework behind those signals:
Pattern 1: You are absent entirely
This is the easiest pattern to recognise.
You run 24 tests.
Competitors appear repeatedly.
Your company appears zero times.
That does not automatically mean your website is poor.
The problem might be:
- weak category association
- unclear service positioning
- little third-party corroboration
- crawler-access problems
- inconsistent company information
- thin sector evidence
- stronger competitors
Start with diagnosis.
Do not respond by publishing twenty new blogs.
Internal link: /ranked-not-recommended/
The first question should be:
What evidence appears to support the companies that are being named?
Pattern 2: You appear occasionally
This result is often more interesting.
Perhaps your company appears for:
commercial cleaning London
but disappears for:
multi-site cleaning contractor for national office estates
Or you appear in broad category questions but not in sector-specific prompts.
That can indicate that the system can identify the business but the association with a particular:
- service
- buyer
- industry
- geography
- capability
is weaker.
That gives you a much more specific content and authority problem to investigate.
You do not need “more visibility” in the abstract.
You may need stronger evidence for one commercial association.
Pattern 3: You appear with outdated or incorrect details
Sometimes being named is not the win.
You may appear with:
- an old company name
- incorrect location
- discontinued service
- wrong sector
- outdated website
- old ownership details
That is an entity-consistency problem.
Look at where the wrong information came from.
If an AI answer provides citations, open them.
Then correct authoritative sources where you legitimately can.
Do not simply edit your homepage and assume the wider web will update immediately.
External records can take longer.
Look at the evidence, not just the names
This is where the test becomes useful.
Do not only write:
Competitor A appeared 14 times.
Record why the answer appears to trust them.
Open the cited sources where the platform provides them.
Look for:
- company website pages
- case studies
- industry publications
- professional bodies
- directories
- review platforms
- news coverage
- framework listings
- government or regulatory sources
The useful question is:
What evidence exists for this competitor that does not exist for us?
You may discover that the competitor does not have dramatically better SEO.
They may simply have clearer digital evidence.
What should you do with the result?
Do not try to copy every competitor.
Turn your findings into three columns.
Already strong
Areas where you appear consistently and accurately.
Protect these.
Weak association
Areas where you appear occasionally.
Investigate the missing evidence.
Completely absent
Important buyer questions where competitors repeatedly appear and you do not.
Prioritise these according to commercial value.
Then look at the six wider levers:
- business information consistency
- extractable content
- external mentions
- reviews where relevant
- crawlability
- entity clarity
If you need a broader assessment rather than manually testing each query:
Do not confuse visibility with market share
This test has limitations.
It does not tell you:
- who has the most customers
- who has the highest revenue
- who will win a tender
- who has the strongest operations
- who is objectively the best company
It tells you which businesses were named by four assistants for the prompts you chose.
That is a visibility signal.
Nothing more.
But for a business already investing in SEO and content, that is an important signal to know.
Because if procurement, prospects or advisers increasingly use assistants during research, being repeatedly absent is at least worth investigating.
Frequently Asked Questions About AI Competitor Visibility
Common questions about repeatability, personalisation, competitor mentions and how to track AI visibility without turning a small test into an unsupported score.
Will the answer change each time I ask?
It can.
Results can change because of differences in retrieval, current web information, model updates, location, prompt context and personalisation.
Do not interpret one answer as a permanent ranking.
Record a baseline and look for repeated patterns across several prompts and assistants.
Does my own search history affect the results?
Potentially, depending on the platform and your settings.
For example, Google documents that Gemini can personalise eligible experiences using past Gemini chats, connected services and preferences.
ChatGPT also provides Temporary Chat, which does not use or create memories, making it useful when you want a cleaner test environment.
OpenAI: ChatGPT FAQ and Temporary ChatThe practical rule is to document how you ran the test and keep the setup as consistent as possible between checks.
Should I use incognito mode?
Incognito mode is not a universal solution.
It can reduce browser-level carry-over such as stored cookies when you are signed out, but it does not automatically disable every account-level personalisation feature if you sign back into a service.
For repeatable testing, fresh chats and controlled personalisation settings are more important than the browser window itself.
How often should I repeat the test?
Monthly is usually enough for a commercial visibility baseline.
Weekly testing can create noise unless you are actively making changes and need a closer diagnostic view.
Use the same core prompts each time and record:
- Date
- Assistant
- Prompt
- Companies named
- Citations
- Your presence
- Accuracy of your information
That gives you something comparable over time.
Should I keep changing my prompts?
Keep a stable core set.
If you rewrite every prompt each month, you lose the ability to compare results.
You can add new questions as buyer behaviour changes, but keep the original benchmark prompts intact.
For a more complete prompt-tracking methodology:
How to build a prompt set for AI visibility trackingIf a competitor appears more often, does that mean they have better GEO?
Not necessarily.
The result only tells you that they were named more often in your test.
There could be several reasons, including:
- Stronger brand recognition
- Better third-party evidence
- Clearer service positioning
- More relevant web content
- Stronger entity signals
- Different source coverage
Use the result to investigate.
Do not turn it into an unsupported GEO score.
Run the test yourself
The point of this exercise is not to create another complicated marketing report.
It should take around ten minutes to begin.
Choose six real buyer questions.
Open:
- ChatGPT
- Gemini
- Claude
- Perplexity
Run the same questions.
Write down every company that gets named.
Then ask:
Who appears repeatedly?
Where do we appear?
Where are we absent?
What evidence supports the companies being named?
That is enough to reveal whether there is something worth investigating.
You may find that your business already appears consistently.
Good.
You may find that you only appear for branded or broad queries.
That gives you a direction.
Or you may discover that three competitors are repeatedly being recommended while your company, despite strong Google rankings, is absent.
That is exactly why the campaign is called:
To run a faster initial check:
Run the free AI visibility check
For wider GEO support:
