How to Find the Buyer Questions Where Competitors Show Up in AI Answers and Your Brand Doesn't

AI visibility gaps are buyer questions where AI assistants name your competitors and not you. This guide outlines a repeatable way to test across major AI platforms, confirm real gaps through repeated runs, and prioritize the ones that matter most commercially.

Here's the direct answer: you find these questions by building a list of unbranded, buyer-phrased questions, running each one multiple times across ChatGPT, Perplexity, Gemini, Claude, and Google AI Overviews, and logging three things for every run — whether your brand was named, which competitors were named, and which sources the answer cited. Any question where competitors appear consistently and you don't is a candidate visibility gap. Then, and this is the part almost everyone skips, you prioritize those gaps by commercial relevance, because a list of missing mentions is homework, not a decision.

This matters most for leaders who already invest in SEO and content, rank well in traditional search, and now suspect there's a blind spot in how AI assistants describe their category. If that's you, this article gives you a method your team can run this week, an honest explanation of where manual testing stops being reliable, and a framework for deciding which gaps deserve budget.

What an AI Visibility Gap Actually Is

An AI visibility gap is a buyer question where an AI assistant names one or more competitors and does not name you.

That definition sounds simple, but it hides a structural difference from traditional search that most executives haven't internalized yet. In classic Google rankings, your page usually still appears somewhere — maybe position eight instead of position two. Generative answers don't work that way. A 2026 arXiv preprint on measuring AI search visibility frames it as a shift from ranking to inclusion: traditional SEO results can move up or down but generally stay somewhere in the list, whereas generative answers tend to either include a brand or leave it out entirely, with little in between from one response to the next. You're either in the recommendation set the buyer walks away with, or you're not. This is exactly the dynamic behind why AI search engines recommend competitors instead of you. Worth noting: this is a preprint (not yet peer-reviewed), and its lead author is affiliated with a commercial AI-visibility firm — treat it as strong, disclosed evidence rather than settled consensus.

The underlying discipline has a name — Generative Engine Optimization — formally defined in academic research accepted to KDD 2024, which also offers a useful dose of honesty: because generative engines are largely opaque and change quickly, brands have limited ability to control exactly when or how their content surfaces in AI-generated answers. That's exactly why diagnosis has to come before action. You can't fix what you haven't observed, and you shouldn't spend against a gap you haven't confirmed. For a closer look at how these systems evaluate brands in the first place, see what AI search engines actually look for.

Three Different Problems Hide Behind "We're Not Showing Up"

Before you test anything, get your vocabulary straight, because three distinct failures get lumped together under "AI doesn't mention us" — and each one has a different fix and a different cost.

• Mention gap: The AI doesn't name your brand when asked an unbranded buyer question. Competitors get named. You don't.
• Source gap: The AI isn't drawing on pages where your brand appears. Even if the model "knows" you exist, the sources it cites for this question don't include you.
• Description gap: The AI names you but frames you wrong — wrong specialty, wrong market segment, wrong price tier, outdated positioning.

We treat this three-layer split as Discovery Authority's working vocabulary rather than an official industry standard, but it's grounded in observation: mention presence and source presence genuinely behave differently. The same 2026 arXiv study tracked how much these day-to-day results actually shifted and found brand mentions held comparatively steadier than cited sources, with source overlap between consecutive days running roughly a third to two-fifths — meaning a clear majority of cited sources turned over daily. Those figures are specific to that study's four-engine, ~45-day observation window, but the pattern they point to is the useful part: mentions and sources are different layers, they move at different speeds, and they should be measured separately. This distinction is also why an AI visibility audit tells you something different than brand mention tracking alone. Lumping them together is the fastest way to misallocate budget — commissioning content to fix what's actually a description problem, or chasing citations when the real issue is that nobody mentions you at all.

The Method: How to Surface Competitor-Present, Brand-Absent Questions

Step 1: Build a buyer-question portfolio, not a keyword list

Start with the questions your buyers actually ask when they don't know you exist yet. These are unbranded and buyer-shaped, not keyword fragments. Useful sources for real buyer language include:

• Sales call recordings and notes — the questions prospects ask in the first ten minutes
• Support and intake tickets — how customers phrase problems before they've adopted your vocabulary
• Your site search logs and the questions your team answers repeatedly
• The "People Also Ask" boxes and related questions around your commercial keywords

Shape those into question types that mirror a buying journey: category questions ("who are the best commercial HVAC service companies in Texas"), problem questions ("why does my ERP implementation keep stalling"), comparison questions ("X versus Y for mid-market firms"), and constraint questions ("best option for a company under 200 employees"). Never include your brand name in the prompt — you're testing whether AI surfaces you unprompted, which is the whole point. If you're not sure where to start, finding the buyer questions your business isn't answering is a useful first pass before you move into AI-specific testing.

Build a portfolio of questions, not one or two. The same 2026 arXiv study looked at how much answers varied prompt to prompt and found the range was wide — some prompts came back nearly identical run after run, while others barely resembled themselves from one attempt to the next — and concluded that testing a single prompt and calling it a visibility reading is unreliable. One question tells you almost nothing. A portfolio tells you a pattern.

Step 2: Test across platforms, because they legitimately differ

Run each question across the assistants your buyers plausibly use: ChatGPT, Perplexity, Gemini, Claude, and Google AI Overviews. This isn't busywork — the platforms differ in documented, mechanical ways:

• OpenAI's own documentation explains that ChatGPT weighs several factors when ranking search results to surface relevant, trustworthy information, that no placement is guaranteed, and that the assistant may pull in live web results automatically when a question calls for current information.
• Perplexity's documentation describes real-time web searching paired with numbered citations attached directly to each answer, linking back to original sources — which makes it especially useful for the source-gap layer, because you can see exactly which pages an answer leaned on. Perplexity also draws on different underlying language models, and lets users choose which one is applied to an answer, so even the same question on the same platform can produce different output.

A brand can be present on one surface and absent on another. That's not a bug in your testing. That's a finding.

Step 3: Run each question more than once — here's why

AI answers are probabilistic. Asking once and concluding "we're invisible" is like judging your whole round off one hole. The same 2026 arXiv study — which tracked four AI engines (ChatGPT, Perplexity, Gemini, and Google AI Mode) daily over roughly 45 days — found that source citations shifted substantially day to day, with overlap holding in a rough one-third to two-fifths range, while brand mentions moved somewhat less but still shifted meaningfully. The authors' core conclusion: visibility should be characterized as a distribution across repeated runs, not a single-point observation.

That study's bootstrap analysis found the standard error of estimated per-brand detection rate dropped below 0.10 at seven runs and below 0.08 at eight runs, and the authors recommend at least seven runs per prompt per day for brand-level monitoring — eight when source-level coverage matters — aggregated on a rolling basis over two to four weeks. One honest caveat before you carve that number into stone: this is a preprint, not peer-reviewed work, the lead author is affiliated with a commercial AI-visibility firm, and the study covered Swiss-German consumer verticals over one window in early 2026. Treat "seven runs" as one study's evidence-based recommendation, not an industry law. The durable principle is simpler: repeat your runs, look for patterns, and never act on a single observation. Even Pew Research, in an entirely separate context, notes that the AI summary for a given search may change over time — a second, independent signal that these surfaces don't sit still.

Step 4: Log the right things — the Gap Confirmation Standard

A gap is only real when it's confirmed across repeated runs, and it's only useful when your log captures enough to diagnose it. Here's the standard we'd hand any marketing team running this internally:

Field	What to record	Why it matters
Buyer question	Exact unbranded prompt used	Makes the test repeatable
Platform and date	Which assistant, which run, when	Answers vary by surface and over time
Your brand named?	Yes / no, and how described if yes	Separates mention gaps from description gaps
Competitors named	Which ones, in what framing	Reveals who currently owns the answer
Sources cited	Which URLs and third parties the answer leaned on	Diagnoses source gaps and points to the fix path
Consistency across runs	How often the pattern repeated (aim for at least several runs per question, per platform)	Distinguishes a real gap from noise

A question earns a place on your confirmed-gap list when competitors appear consistently across repeated runs and your brand consistently doesn't. A one-time absence is noise. A pattern is evidence.

Step 5: Rule out the cheap fix first

When a competitor appears and you don't, the cause sits in one of three places: the model isn't reaching pages that mention you, the pages it does reach describe your category without you in it — or it can't access your site at all. Rule out that last one first, because it's the cheapest check you'll ever run. OpenAI documents that to make a website eligible for inclusion in ChatGPT search, site owners must allow OAI-SearchBot to crawl the site and confirm that the website host or CDN allows traffic from OpenAI's published searchbot IP addresses. If your robots.txt or CDN is blocking AI search crawlers, no amount of content will close the gap. Check access before you commission anything.

Why a Confirmed Gap Isn't Automatically a Priority

This is where every generic guide — and frankly, most AI-generated answers on this topic — stops. They hand you detection mechanics and a spreadsheet of gaps, then leave you alone with it. But a Founder or CMO doesn't need a spreadsheet. They need to know which few gaps are worth fixing this quarter and which can wait.

Here's the prioritization framework we use at Discovery Authority — presented as our point of view, not an industry standard. Score each confirmed gap against four plain-language questions:

1. Revenue proximity. Is this a question a buyer asks near a purchase decision, or three steps upstream? "Best commercial roofing contractor for multi-site property managers" is close to money. "How do roofs work" is not.
2. Competitive concentration. Is one rival owning the answer across platforms and runs, or is the answer fragmented among many names? Fragmented fields are typically easier to enter.
3. Gap type. Mention, source, or description? A description gap might be fixable with clearer positioning on pages you already control. A source gap may require third-party presence you don't control, which takes longer.
4. Fixability. Can you close this with owned content and site changes, or does it depend on external citations, reviews, and time?

Where you start is the intersection: revenue-proximate, fragmented, and fixable with assets you control. Real gaps that are upstream, dominated by a single entrenched competitor, and dependent on third-party sources are still real — they're just a next-quarter problem, not a this-quarter one. Notably, the KDD 2024 GEO research found that the efficacy of optimization strategies varies across domains, underscoring the need for domain-specific approaches over blanket fixes.

No multiplication formula, no fake-precision score. A grid and senior judgment beat an equation dressed up as math.

Why Being Named in the Answer Matters Even When Nobody Clicks

The natural executive question is "fine, but how much traffic does this drive?" — and the honest answer is often "very little," which is exactly the point. Pew Research found that when a Google AI summary appeared, users clicked a link inside the summary in just 1% of visits, and were more likely to end their browsing session entirely after seeing one (26% versus 16% without a summary). In the same study, about 58% of participants encountered at least one AI summary in a single month of searching (March 2025).

Two important caveats: Pew's data covers Google only, based on browsing activity from 900 U.S. adults during March 2025, and Pew itself notes the AI summary for a given search may change over time — don't stretch those exact numbers to ChatGPT or Perplexity behavior. But the strategic implication travels well: on AI surfaces, the click is increasingly not the asset. The asset is whether your name is in the shortlist the buyer walks away with. That's what makes mention gaps commercially real, not scary — and it's why this work rewards calm evidence-gathering rather than panic, as we've argued in AI search is deciding who gets the call.

Where Manual Testing Stops Being Enough

Everything above can be run manually by a capable marketer, and we'd encourage you to try it — even a rough version will teach you more about your AI visibility than any think piece. But be honest about where the DIY version strains:

• Volume. A meaningful question portfolio, across five platforms, with repeated runs, over a multi-week window, is hundreds of logged observations. That's a real workload landing on a team that's already stretched.
• Consistency. The value of the data depends on disciplined, repeatable methodology. Ad hoc testing produces ad hoc conclusions.
• Interpretation. Knowing that a source gap exists is different from knowing which cited pages matter, why competitors keep appearing in them, and what the realistic fix path is.
• Prioritization. Connecting gaps to revenue-relevant buyer questions — for your services, your geographies, your deal sizes — is a judgment call, not a filter setting.

This is exactly what Discovery Authority's AI Search Visibility Audit is built for: a structured, competitive analysis of observed visibility across Claude, ChatGPT, Perplexity, Gemini, and Google AI Overviews, mapped against the buyer questions that matter commercially for your business, delivered as a prioritized roadmap focused on high-impact gaps first rather than a wall of raw findings. If you want the details behind how that scoring works, see how NarraLoom scores AI search visibility. One thing we're direct about: every finding is a point-in-time snapshot of observed behavior on specific surfaces. These systems change, and nobody — including us — controls how they rank, cite, or recommend brands. What an audit gives you is evidence and priorities instead of guesswork, which is what a leadership team actually needs to make a funding decision.

And when confirmed gaps point to a content problem, that's where the Content Authority System picks up — consistent, human-reviewed articles and branded LinkedIn and X content built against evidence-backed gaps rather than generic keyword lists, designed to be more operationally efficient than labor-heavy traditional production. For clients whose gaps also involve technical SEO or paid search, our Search & Paid Media services — delivered in partnership with Adwest — address that layer directly.

Frequently Asked Questions

How is AI visibility measured?

Honestly measured, it's a set of observations with a disclosed method — not a single score. Meaningful measurement states how many buyer questions were tested, on which platforms, how many times each, over what time window, and what was recorded: whether you were named, which competitors were named, which sources were cited, and how you were described. A 2026 arXiv study of AI search measurement recommends characterizing visibility as a distribution rather than a single-point outcome, aggregated on a rolling basis over two to four weeks. Be skeptical of any dashboard reporting one "visibility score" without disclosing its run count and time window — that's a distribution compressed into a point estimate.

Which AI platforms should we monitor?

Start with the five with the broadest buyer reach: ChatGPT, Perplexity, Gemini, Claude, and Google AI Overviews. They differ mechanically — ChatGPT's placement is explicitly not guaranteed per OpenAI's documentation, and Perplexity attaches numbered live-web citations to every answer, which makes source analysis easier. Weight your attention by where your buyers actually are: consumer home-services buyers may lean more on Google's AI experiences, while B2B evaluators increasingly research in ChatGPT and Perplexity. Presence on one platform does not predict presence on another, which is why single-platform monitoring gives a false sense of security in either direction.

Is this a one-time audit or an ongoing practice?

The evidence says AI answers can shift materially over days and weeks, so a single audit is a snapshot — genuinely useful for finding and prioritizing gaps, but not a permanent record. A sensible rhythm is a thorough baseline audit, targeted remediation of priority gaps, then periodic re-measurement to see whether patterns changed. Treat it like competitive intelligence, not a certificate you hang on the wall.

What do we do after we find the gaps?

It depends on gap type. Description gaps often mean clarifying positioning on pages you control. Mention gaps usually mean building clear, authoritative content that directly answers the buyer questions where you're absent. Source gaps typically require presence in the third-party pages AI answers cite — a slower, steadier build. And access issues — crawler blocks, CDN restrictions — should be ruled out first because they're cheap to fix and undermine everything else. The right sequence comes from prioritization, not from fixing everything at once.

Start With Evidence, Not Anxiety

The question "where do competitors show up in AI answers when we don't?" is answerable — rigorously, this quarter, without hype and without panic. Build a real buyer-question portfolio, test across platforms, repeat your runs, log mention, source, and description layers separately, and then apply commercial judgment to decide what's worth fixing first. That's the difference between a stack of screenshots and a strategy.

If you'd rather have a senior team run the full structured version — competitive evidence across Claude, ChatGPT, Perplexity, Gemini, and Google AI Overviews, tied to the buyer questions that matter for your business, with a prioritized roadmap instead of a raw gap list — that's exactly what the AI Search Visibility Audit does.

Call Discovery Authority to discuss at 925-963-5767 or click to schedule time.

---