Find out whether AI platforms recommend your firm, or your competitors instead.

What is an AI visibility report?

An AI visibility report measures whether and how a company appears when prospective customers query AI platforms for recommendations. Rather than simply counting brand mentions, the report examines commercially relevant buyer questions, identifies which competitors are being recommended instead, surfaces the sources and signals associated with those recommendations, and identifies practical opportunities to improve visibility.

An AI visibility report is much more than simply checking if ChatGPT knows your company exists.

What questions does an AI visibility report answer?

First, not all AI visibility reports are equal. There is no approved standard, set format or even accepted terminology. You may see similar analysis described as an AI Search Audit, LLM Visibility Report, Generative Engine Optimization (GEO) Report, AI Share of Voice (SOV) Report or Answer Engine Optimization (AEO) Audit. The terminology varies, but the more important question is what the analysis actually tells you.

At the simplest end of the market, an AI visibility report can be little more than an automated snapshot: enter a domain, run a predefined set of prompts and receive a visibility score, a list of brand mentions and perhaps an AI Share of Voice (SOV) comparison with a few known competitors.

That can be useful for a quick check. But it doesn't necessarily tell you whether the right questions were tested, whether AI is recommending your company rather than simply mentioning it, which competitors you didn't know about are being surfaced, or why those companies are appearing instead of you.

A useful AI visibility report should go beyond "How visible are we?" and answer the more important questions that follow from it.

Does AI recommend my company when buyers are looking for businesses like mine?

An AI platform can be well aware your company exists and still never recommend it. The commercially important question is whether your company appears naturally when a prospective customer asks for recommendations around the services, problems, location and other requirements you can genuinely fulfil.

A good AI Search Audit or LLM Visibility Report therefore needs to examine the buying conversations that matter, rather than simply checking whether your brand can be made to appear in an AI response.

Which buyer questions do we appear for, and where are we missing?

AI visibility is rarely all or nothing.

A company might be regularly recommended for broad category searches but disappear when the buyer adds a particular industry, technical requirement, regulation, location or use case.

A useful report should identify these differences across relevant buying conversations, including:

• Category searches
• Use cases
• Comparisons
• Pricing
• Reviews and online reputation
• Compliance requirements
• Specialist capabilities
• Implementation and switching questions

These are the actual AI visibility gaps: situations where your company could reasonably be considered, but isn't.

Which competitors are AI platforms recommending instead?

A visibility report shouldn't only compare you with the competitors you already know.

One of the most valuable findings can be discovering companies that AI systems repeatedly recommend alongside, or instead of, your business.

These may include established competitors, niche specialists or companies you haven't previously considered significant.

Your AI competitors are the companies being put in front of your prospective customers, whether or not they were on your original competitor list.

Are we being mentioned, or actually recommended?

A brand mention isn't necessarily a recommendation.

Your company could appear as background information, in a citation, as part of a long list, or even in a negative context. None of those is equivalent to an AI assistant presenting the company as a suitable choice for the buyer.

This is one of the limitations of treating an AI Share of Voice (SOV) Report as the complete picture. Share of voice can be a useful metric, but a meaningful visibility analysis should also examine the context and prominence of appearances, not simply count how many times a company name occurs.

How does our visibility differ between AI platforms?

There is no single set of "AI search results". AI-generated answers differ significantly between platforms. ChatGPT, Gemini, Claude, Perplexity, Copilot and other AI search experiences can return different companies for similar buyer questions. A business that appears frequently on one platform may be almost invisible on another.

A useful LLM Visibility Report should therefore identify platform visibility gaps, rather than treating the output of one AI system as representative of the entire AI search landscape.

Why are competitors appearing for AI queries when we aren't?

Finding that a competitor was recommended 18 times while your company appeared four times tells you what happened. It doesn't tell you what to do about it.

A detailed analysis should investigate the differences between the companies being recommended and those being overlooked. That can include service and use-case content, positioning, case studies, reviews, specialist expertise, industry associations, directories, regulatory evidence and the third-party sources associated with AI responses.

This is also where a genuinely useful Generative Engine Optimization (GEO) Report or Answer Engine Optimization (AEO) Audit needs to move beyond generic optimisation recommendations. Before deciding what to change, you need evidence about where the visibility gap actually occurs and what distinguishes the companies currently being recommended.

AI platforms don't expose a complete account of why a particular company was selected. What's checkable is the observable gap: the differences and evidence that could reasonably explain why one company is easier for an AI system to identify, verify and recommend than another.

Which authority sources and third-party websites matter in our market?

Your own website is only part of your AI visibility.

AI responses can draw upon industry publications, directories, trade bodies, professional associations, regulatory sources, review platforms, comparison sites and other independent online sources.

A strong AI Search Audit should therefore examine the sources appearing around your market and competitors, rather than limiting the analysis to your own website.

This can reveal an entirely different type of visibility gap: competitors may have stronger representation in the external sources AI systems use to research or support their answers, while your company is largely absent from them.

That is a problem no amount of rewriting your homepage alone will solve.

What should we actually do about the gaps?

A report containing hundreds of prompts, charts, mentions, percentages and AI Share of Voice scores may look impressive while leaving the recipient with exactly the same question they had before:

What do we need to change?

A good report should connect the evidence:

Some gaps may require clearer service or use-case content. Others may point towards weak third-party corroboration, missing industry listings, inadequate proof of specialist expertise, unclear positioning or simply a buying conversation the company isn't currently well equipped to compete for.

The output should be a prioritised set of findings and actions, not simply another dashboard.

The difference between reporting and diagnosis

Whether it's labelled an AI Visibility Report, AI Search Audit, LLM Visibility Report, GEO Report, AI Share of Voice Report or AEO Audit, the name matters less than the depth of the analysis behind it.

Automated visibility reporting, and tracking AI visibility over time once you know which prompts, competitors and platforms matter, can be extremely useful. But a detailed AI visibility report has a different job.

It should help establish what should be measured in the first place, discover where the important gaps are, investigate why they exist and determine which of them are worth addressing.

A basic AI visibility report tells you where you appeared.

A good AI visibility report tells you where you should have appeared, who appeared instead, what evidence may explain the difference, and what you can do next.

We don't start by asking you for a list of prompts

Automated AI visibility tools typically ask for a domain and a handful of prompts, then run them and hand back a score. That skips the hardest part of the whole exercise.

Most companies don't have a definitive list of the questions prospective customers ask ChatGPT, Gemini, Claude or Perplexity when looking for businesses like theirs. Ask a business to supply fifty prompts and what usually comes back is fifty versions of how they'd like to be asked about, phrased in their own marketing language rather than how a buyer with no relationship to the company actually types the question.

That gap matters. A prompt list built from assumptions bakes those same assumptions into the analysis before testing even starts.

The starting point here is the market and the buying journey itself. Before any prompt gets tested, the plausible commercial conversations happening around whatever you sell need mapping in detail, across areas such as:

• Category
• Use case / fit
• Comparison
• Pricing
• Reviews / online reputation
• Compliance
• Specialist requirements
• Implementation
• Switching / replacement
• Location
• Problems / outcomes

Prompts get built against that map, not the other way round.

Most people, asked to supply prompts, default to typing the way they'd type into Google. A typical Google search runs to three or four words. A real conversational question to ChatGPT or Perplexity runs far longer, often dozens of words, with several requirements folded into one sentence rather than typed as separate searches.

Content built to answer a four-word keyword doesn't automatically answer the fifteen different longer, more specific versions of that same question a buyer might actually ask. That's the design problem the mapping above solves: build against the conversation, and the short-form and long-form versions of the same query tend to get covered without needing to write for each one separately.

We don't start by assuming who your AI competitors are either

That's commercially valuable information. A predetermined competitor list only tells you how you compare against the businesses you already had in mind. It says nothing about the ones AI is putting in front of your prospects instead.

The approach here is to record the companies AI actually surfaces across the buying conversations mapped earlier, rather than restricting the analysis to a list supplied up front. Whoever gets named, gets recorded, known rival or not.

There's usually a reason those unexpected names appear, and it's rarely about the quality of their work. AI platforms lean heavily on third-party sources when building a recommendation, not just a company's own website. In most niches, a handful of trusted authority sites carry disproportionate weight: trade directories, industry bodies, professional associations, comparison and review platforms specific to that sector. Press and media coverage adds another layer, an interview, a feature, a quoted comment in a trade publication. Social media activity and community discussion (LinkedIn, Reddit, niche forums) can carry weight too, particularly on platforms like Perplexity that pull heavily from community sources.

A company that's well represented across those sources tends to get recommended more often than a company with a better service but a thinner footprint outside its own site. That's often the real explanation behind an unfamiliar name showing up: not that the company is secretly better known, but that it's better corroborated in the places AI is actually looking.

My AI visibility report vs a generic automated AI visibility tool

A generic automated tool typically:

• Runs a templated set of prompts against your domain
• Scores you against competitors it guessed at from a quick search
• Has no idea which capability, region or buyer type actually matters to your business
• Can read a genuine strength as a weakness, or miss it entirely, simply because it has no context for what it's looking at

This report works differently, because it starts with a conversation, not a domain field:

• The buyer-conversation map gets built from what you tell us you actually want to win, not from a template
• The competitors recorded are the ones AI genuinely surfaces for those conversations, not a guessed shortlist
• Findings get checked against what actually matters to your business before anything gets prioritised
• Nothing goes into the report that a five-minute automated scan would have gotten wrong or missed

That's also why the report isn't produced from a website alone. A short discussion first means the analysis is built around your actual aims, not a best guess at them. It's the difference between a report that tells you something true but generic, and one that tells you something specific enough to act on.

A short conversation often surfaces the actual objective on both sides too. Not just for the client: sometimes what a business genuinely needs only becomes clear once someone asks the right question.

AI visibility report vs an SEO report

An SEO report asks whether a page can be found and ranked: does it target the right keywords, does it satisfy technical crawl requirements, does it earn backlinks, does it outrank competitors on a results page. The output is a ranking position, a set of keywords, a technical fix list.

An AI visibility report asks a related but separate question: when a prospective customer asks an AI system to recommend or compare providers, does the company get named, and named as a genuine recommendation rather than background noise. The output is a gap map across real buying conversations, a competitor list built from what AI actually surfaces, and a set of findings tracing back to specific evidence.

The two overlap more than they compete. Generative Engine Optimization, the practice of improving AI visibility specifically, shares real ground with SEO: both reward clear, well-structured content and genuine third-party corroboration. But a page can rank well on Google and still never get cited in an AI answer, and the reverse happens too. Ranking position and AI citation are measuring different things, built from different signals, checked by different platforms.

A business with strong SEO and no AI visibility work has usually never had this side of it checked at all. The two aren't substitutes for each other. A company can be doing well on one and be genuinely invisible on the other, and won't know it without looking, plus won't know why without a proper analysis.

What your AI visibility report includes

Everything above is the groundwork. This is what it turns into.

1. Buyer conversation analysis

Identifies the commercially relevant questions around the client's category. This isn't conventional keyword research. It's identifying situations in which a prospective customer could reasonably ask an AI system to recommend or compare providers.

2. AI platform testing

The mapped conversations get run across the AI platforms that actually matter for that market, not just one. ChatGPT, Gemini, Claude, Perplexity and Copilot can return different companies, and different AI-generated answers altogether, for the same question, so testing one and calling it "AI search" misses most of the picture.

3. Brand recommendation visibility

The report checks whether the subject appears as a brand mention, and just as importantly, whether that mention rises to an actual recommendation, or reads as neutral, favourable, or actively cautioning against the business. That's a qualitative read of how AI portrays the business, not a scored metric, but it's often as commercially telling as whether the business appears at all. Where the data supports it, possible data points include:

• Recommendation frequency
• Citation rate across platforms
• Prompt coverage
• Platform coverage
• Recommendation position and prominence
• Category visibility
• Use-case visibility

Only metrics genuinely calculated for that engagement get reported. No padding a page with numbers that weren't actually measured.

4. AI competitor discovery

Every company that actually appears in a response gets recorded, not just the ones on a predetermined competitor list. From there: who appears most often, for which conversations, on which platforms, who consistently outappears the subject. That builds real competitive marketing intelligence and starts to uncover the specific prompts competitors are winning that the subject currently isn't, often a different list to the one a client would have written down themselves.

5. AI visibility competitor gap analysis

Discovery turns into gaps once it's laid out side by side. A simplified example of what that looks like:

Buyer conversation	Your company	Competitor A	Competitor B
Category recommendation	✓	✓	✓
Specialist requirement	—	✓	✓
Compliance requirement	—	✓	—
Comparison query	✓	✓	✓
Implementation requirement	—	✓	✓

The real report carries more detail than this, but the shape is the same: a gap is a conversation where a competitor shows up and the subject doesn't, or where the brand isn't included in the answer at all.

6. Citation and source analysis

This is where a proper report earns its keep over an automated tool. Rather than stopping at "Competitor A appears more," the analysis assesses what surrounds that visibility, including:

• Competitor websites
• Industry publications
• Trade associations
• Directories
• Review platforms
• Comparison sites
• Regulatory sources
• Professional bodies
• News and media
• Specialist blogs
• Case studies
• Partner websites

AI platforms are generative AI engines built on large language models. A brand's presence in a model's training data is only part of the picture. Many platforms now use retrieval-augmented generation too, pulling answers from indexed online sources and third-party content in addition to what the model already learned. That's the mechanical reason third-party sources carry so much weight in what gets recommended.

Run this across enough companies and the same handful of source types keep showing up around the ones getting recommended. They're rarely the ones a client expected. It's also worth noting which competitors, and which terms, AI associates a business's name with, since that pattern of association often says as much about positioning as the citation itself does.

7. Why competitors may be winning

This is the diagnostic core of the report. It looks at:

• Positioning: does the competitor explicitly describe itself as suitable for the requirement?
• Content: does it have a dedicated page for the capability, where the subject has one sentence on a generic services page?
• Third-party corroboration: do independent sources associate the competitor with that capability?
• Authority: is the competitor listed in relevant industry or trade sources?
• Reviews and reputation: is there external evidence supporting suitability?
• Entity clarity: can AI systems clearly establish what the company does, where it operates and who it serves?
• Specialism: is there stronger evidence around the precise niche?

Together, these are the signals that earn AI recommendations, not any single factor in isolation. They map onto the same three checks the report runs against every gap: consensus alignment, entity depth and citable structure.

One thing that shows up again and again running these checks: a well-written page can still go uncited simply because it restates the same view already sitting elsewhere on the web, rather than adding something an AI model hasn't already encountered from another source. The pages that get picked up tend to have something genuinely original in them, a specific case, an unusual angle, a real number nobody else has published, not just competent writing saying the same thing everyone else already says. Good prose alone has never been enough to explain what gets cited and what doesn't.

None of this proves causation on its own. It's evidence and a plausible explanation, stated as exactly that.

8. Third-party visibility gaps

Some gaps sit away from the client's own website entirely. Competitors listed in an industry directory the subject isn't in. Competitors with trade-body profiles the subject doesn't have. Competitors named in independent comparison articles the subject is absent from. That split, owned-source gaps against third-party-source gaps, changes what the fix actually looks like.

9. Website and content gaps

Then the parts the client does control. Findings here can include:

• Missing service pages
• Weak category definition
• Vague positioning
• Thin specialist content
• Missing case studies or evidence
• Weak location or category association
• Insufficient compliance information
• Unclear implementation detail
• No FAQ content addressing real buyer concerns

None of this becomes generic SEO advice. Every recommendation traces back through the same chain: finding → evidence → visibility gap → recommended action.

10. Prioritised action plan

Not a list of eighty-seven things that could be improved. Three tiers instead:

Priority 1: highest-impact gaps. Directly tied to the buying conversations where competitors currently dominate.

Priority 2: supporting improvements. Changes that strengthen overall entity and category understanding.

Priority 3: longer-term opportunities. Third-party mentions, authority building, citations.

That's what makes the report something to act on, not just read.

Example: from AI visibility gap to action

An illustrative example, not a real client, to show the shape of it.

A boutique planning consultancy asks ChatGPT and Perplexity a version of the same question a prospective client might: who handles retrospective planning applications for listed buildings in their region. Their own firm doesn't appear in either answer. Two other names do, one a larger regional practice, one a sole practitioner nobody at the firm had previously considered a competitor.

Checking the sources behind those two answers turns up a pattern. Both named firms have profiles on a regional planning directory the consultancy has never listed with, and both have been quoted in local press coverage of listed-building disputes. The consultancy's own website has a strong page on the exact service, better written, more detailed, than either competitor's. It just isn't showing up anywhere AI platforms are actually looking for corroboration.

• Buyer conversation: retrospective listed-building planning applications, specific region
• Visibility gap: the firm doesn't appear, two competitors do
• Competitor advantage: directory listing plus press coverage neither the firm nor most of its real competitors have
• Recommended action: get listed on the directory, pursue the same kind of press angle that got the other two firms quoted

Two of those, the directory listing and a press mention, are things a website rewrite alone was never going to fix. The service page didn't need work. The problem was never on the site at all.

That last part, closing the gap once it's been found and diagnosed, is where the actual work happens. There's more detail on how a visibility gap becomes a diagnostic finding and then an executed fix elsewhere on this site.

Who needs an AI visibility report?

Some typical situations worth recognising, before any jargon gets attached to them:

• A competitor keeps coming up in conversations with prospects, and it's not always clear why, since the work itself isn't obviously better
• Referrals and word-of-mouth are steady, but new enquiries from anyone who hasn't heard of the business directly have quietly dried up
• A genuine specialism exists, something the business is actually better at than most competitors, but it never seems to get found by anyone searching for exactly that
• Someone's mentioned that ChatGPT or a similar tool didn't recommend the business when they tried asking it a question the business should have won
• The website gets decent traffic, or used to, but enquiries haven't kept pace with it

None of these need the words "AI visibility" attached to notice them. They're everyday complaints from businesses losing ground and they can't explain why.

What often connects them is same problem: a prospective customer asked an AI platform a question this business could genuinely answer well, and got sent to someone else instead, without either side knowing it happened. It's the business going invisible at exactly the decision-making moment, not through any fault in the actual work.

This tends to matter most for small businesses, boutique firms and specialists who compete on genuine expertise rather than marketing budget, since a large, well-known competitor keeps showing up regardless, and a strong, lesser-known specialist has the most to gain from being found accurately. A business already dominant in its category, with the kind of scale and name recognition that gets it recommended by default, has less riding on this than one that wins on the strength of the actual work.

What an AI visibility report does NOT tell you

The AI visibility and GEO space is relatively new but has picked up some bad habits fast.

The GEO/AEO services niche is currently flooded with hype. Here is what we won't falsely promise you:

• Guaranteed citations. No credible analysis can promise a specific AI platform will recommend a business by a set date. Platforms change how they retrieve and rank sources constantly, sometimes without any public announcement.
• A single "AI SEO score" that means the same thing everywhere. There's no industry standard for this. A score from one tool isn't comparable to a score from another, and neither tells you much on its own without the underlying evidence behind it.
• Proof that a specific fix caused a specific result. If visibility improves after changes get made, that's a good sign, but AI platforms don't publish their reasoning, and other things change at the same time. Evidence and correlation, not proof of causation.
• Real-time monitoring. This report is a snapshot, built from testing done at a specific point in time, across specific platforms, for specific buyer conversations. It's accurate for that moment. Platforms can change what they surface the following week.

A few genuine limits worth naming too:

• The report tests the buyer conversations identified as commercially relevant during the initial discussion. It can't cover every possible way someone might ask, and a conversation nobody thought to test won't show up in the findings.
• Occasionally, an AI platform states something factually wrong about a business during testing, gets a detail wrong, confuses it with a similarly named competitor, or misattributes a service. That gets flagged when it happens, but checking for it isn't a dedicated audit in itself, and it's a separate problem from visibility, worth knowing about but not something every business will encounter.
• Findings are built from testing across major AI platforms at the time of the report, and AI answers are far less stable than a traditional search results page. A Google ranking tends to hold for weeks or months. What an AI platform cites for the same question can shift from one week to the next, sometimes with no visible cause. A finding accurate today isn't guaranteed to hold indefinitely, which is part of the case for checking again periodically rather than treating one report as settled.

None of this is a case against doing the analysis. It's the difference between a report you can trust because it's honest about what it can't tell you, and one that oversells itself to close a sale.

How the AI visibility analysis works

A walkthrough of the actual process, from first conversation to finished report.

1. A short conversation with the founder or ownerBefore anything gets tested, a conversation covers what the business actually wants to win, its actual customer profile and goals rather than an assumed one, who it competes against in its own mind, and what it believes is true about its market. This is also where any wrong assumptions surface early, both the client's and any that might otherwise get baked into the analysis unchecked.
2. Define the market and the buying journeyThe commercial territory gets mapped properly: what's actually being sold, to whom, and the range of situations where a prospective customer might reasonably ask an AI system for a recommendation.
3. Map the buyer conversationsThat territory gets broken into the specific conversations worth testing: category, use case, comparison, pricing, reviews and reputation, compliance, specialist requirements, implementation, switching, and location, built from what came out of the conversation in step 1, not a generic template. Each conversation also gets tagged by intent: informational, transactional, or commercial investigation, so the prompt set covers genuine research questions as well as ones closer to an actual buying decision, rather than skewing toward one type.
4. Build the actual prompt setPrompts get written in a range of realistic phrasings, not just the short, keyword-style version most people default to when asked for a prompt. A real buyer might type a four-word search or a much longer, multi-part conversational question in the same sitting, so both get tested rather than just the version that's easiest to write.
5. Run the prompts across AI platformsThe full set gets tested across the platforms that matter for that market, typically ChatGPT, Gemini, Claude, Perplexity and Copilot, since results genuinely differ between them.
6. Record what comes backEvery response gets logged: whether the business appears at all, whether it's mentioned or actually recommended, and where it sits in the answer. Every other company named gets recorded too, not just the ones on an assumed competitor list.
7. Build the real competitor landscapePatterns get pulled from that data: who appears most often, for which conversations, on which platforms, and who consistently outappears the business being analysed.
8. Investigate the sources behind the resultsFor the companies getting recommended, the sources associated with those results get checked: directories, trade bodies, press coverage, review platforms, and other third-party sites, since this is usually where the real explanation sits.
9. Map the gapsBuyer conversations, the business, and its competitors get laid out side by side, so it's clear exactly where the business is missing and who's there instead.
10. Diagnose whyEach gap gets checked against three things: whether the business's content matches what AI already treats as the consensus view, whether AI can resolve the business as a real, well-corroborated entity, and whether the site actually has content AI can cleanly cite and prioritise when constructing an answer. That's usually where the real answer to "why them and not us" sits.
11. Build the prioritised action planFindings get sorted into what would close the highest-impact gaps first, what strengthens the business's overall standing more generally, and what's worth pursuing longer-term.
12. Go through the findings togetherThe report gets talked through, not just sent over, so there's room to ask questions and work out what closing these gaps would actually involve.

Frequently asked questions