"Who does ChatGPT think my competitors are?" is fast becoming one of the most useful questions a marketing team can ask—and one you cannot answer from your CRM. The short answer: ChatGPT's version of your competitive set is the group of brands it repeatedly names alongside or instead of you when buyers ask for recommendations. That set is inferred from training data and live retrieval, not from your pipeline. It routinely includes rivals your sales team never logs and omits ones they obsess over. This guide shows you the exact prompts to run, how to score the results with real data, and how to reconcile the AI set with the lists you already trust.
What "who does ChatGPT think my competitors are" actually means
Your AI-native competitive set is the group of brands ChatGPT names in the same breath as yours when a buyer asks for options—ranked by how often it recommends each one. It is not the list of "alternatives to ChatGPT" (a common search mix-up), and it is not your official battlecard. It is a behavioral set: whoever the model treats as your peer in answer after answer.
This distinction matters because buyers now see that set before they see you. When someone asks ChatGPT, Gemini, or Perplexity "what's the best tool for X," the answer is a shortlist. The brands on that shortlist are, functionally, your competitors in the AI channel—whether or not your revenue team agrees. Tracking brand mentions in ChatGPT turns that invisible shortlist into something you can measure and influence.
Why your AI-native competitive set differs from your sales list
AI clusters competitors by language patterns and co-mention, not by deal overlap. Your sales team defines rivals by who you lose deals to. A large language model defines them by which brands appear together across web content, reviews, forums, and comparison pages—then reproduces those clusters in its answers. The two lists overlap, but they are never identical.
Three forces pull them apart:
- Category framing. If the web describes you as a "project tool" and a rival as a "work OS," the model may group you with adjacent products you'd never bid against. When it files you under the wrong header, the whole set skews—a problem worth diagnosing when AI puts your product in the wrong category.
- Incumbent bias. Models lean toward brands with dense, repeated coverage, so household names surface even in niches they barely serve. This tilt is measurable, as our look at whether AI favors big brands shows.
- Retrieval freshness. Live web retrieval can inject a fast-growing startup into your set months before your CRM registers it as a threat.
The result: the AI competitive set is an early-warning system your sales list can't replicate.
How to find out who ChatGPT thinks your competitors are
To discover your AI-native competitive set, run a panel of buyer prompts across models, extract every brand named, and score them by recommendation frequency and co-mention with you. One-off prompting won't do it—answers vary run to run, so you need volume and structure. Here is the repeatable method.
- Build a prompt panel. Collect 30–50 questions real buyers actually type, weighted toward the bottom of the funnel where recommendations happen. Cover four shapes:
- Category: "What's the best [category] tool for [use case / team size]?"
- Alternatives: "[Your brand] alternatives" and "tools like [your brand]"
- Head-to-head: "[Rival A] vs [Rival B]" and "is [your brand] or [rival] better for X?"
- Job-based: "How do I [the job your product does]?"—recommendations sneak in here too.
Want a fast read today? Run five of these prompts five times each; even that thin panel exposes the core set.
2. Run each prompt multiple times, across models. Query ChatGPT, Gemini, Perplexity, Claude, and Copilot—five or more runs each—to average out variance. Different engines pull from different sources, so a brand can top one and be absent in another; running all of them is the only way to see the whole set.
3. Extract every named brand. Parse each answer for brand names, including ones you don't recognize. Don't filter yet.
4. Score by two signals. Track recommendation frequency (share of answers a brand appears in) and co-mention rate (share of your answers that also name that brand). Co-mention is the sharper signal of a true peer.
5. Cluster into tiers. Sort brands into core rivals, adjacent players, and peripheral or emerging names based on the two scores.
6. Reconcile with human lists. Compare the AI set against your sales and market-research rosters, then investigate every mismatch.
Steps 3–4 are the hard part by hand: extracting and scoring brand names consistently across hundreds of answers is where most manual attempts break down. That extraction, scoring, and frequency math is exactly what a purpose-built tracking tool automates.
A worked example: extracting one brand's AI-native competitive set
Here is an anonymized example from our own tracking. We ran a 40-prompt panel, five runs each, across ChatGPT, Gemini, and Perplexity over two weeks—600 total answers—for a mid-market project-management SaaS we'll call Northstar PM. We then measured how often each brand appeared and how often it was co-mentioned with Northstar.
| Brand | Appeared in answers | Co-mentioned with Northstar | Assigned tier |
|---|---|---|---|
| Asana | 78% | 61% | Core rival |
| monday.com | 74% | 58% | Core rival |
| ClickUp | 69% | 55% | Core rival |
| Trello | 52% | 40% | Adjacent |
| Notion | 47% | 33% | Adjacent (category-blur) |
| Smartsheet | 28% | 19% | Peripheral |
| Wrike | 24% | 17% | Peripheral |
| Linear | 21% | 15% | Emerging / indirect |
| Northstar PM | 12% | — | The client |
Two findings stood out. First, Northstar appeared in only 12% of answers—so in 88% of buyer recommendations, its own competitive set showed up and Northstar didn't. That is a share-of-voice gap, not a product gap. Second, Notion surfaced in nearly half of answers despite never appearing on Northstar's sales battlecard, because reviews and listicles kept grouping the two under "flexible work tools." That co-mention density is exactly what a brand co-mention analysis is built to map.
These numbers describe one representative project, not a universal benchmark—your set will differ. The method, however, transfers directly.
Reconciling the AI set with your sales and market-research lists
Once you have the AI-native set, sort every brand into one of three buckets: confirmed rivals, AI-only "phantom" rivals, and sales-only "invisible" rivals. Each bucket implies a different action, and the mismatches are usually worth more than the agreements.
| Segment | What it means | Action |
|---|---|---|
| Confirmed rivals (AI + sales agree) | Your real head-to-head competitors, validated by both signals | Prioritize comparison content and evidence to win the shortlist |
| AI-only phantom rivals | Brands AI keeps naming that your team doesn't track | Investigate positioning and category framing; decide whether to embrace or dispute the grouping |
| Sales-only invisible rivals | Brands you lose deals to that AI never names | Either they're weak in AI search (an opening for you) or you're targeting a segment the model files elsewhere |
The phantom bucket is your positioning radar—it reveals how the model has categorized you. The invisible bucket is your opportunity map: if a genuine rival is missing from AI answers, that channel is still up for grabs. This reconciliation is also where reputation work begins, because the set the model repeats back is, in practice, the reputation buyers meet before they ever reach your site.
Direct vs indirect: classifying the competitors AI surfaces
Not every brand in your AI set competes for the same budget—separate direct rivals from feature-adjacent neighbors before you act. A direct competitor solves the same job for the same buyer; an indirect one shares a use case or a keyword but sits in a different budget line. In the Northstar example, Asana and monday.com are direct; Notion is indirect—a docs-and-notes tool the model blurred into the set.
Why bother splitting them? Because the response differs. You fight direct rivals with head-to-head proof and comparison pages. You handle indirect ones with category clarity—teaching answer engines where you belong so the model stops filing you next to the wrong product. Our guide to separating direct vs indirect competitors in AI answers breaks down how to tell them apart at scale, and why conflating the two wastes budget on the wrong battles.
Getting this classification right is the difference between chasing every name in the set and focusing on the handful that actually move revenue.
What to do once you know your AI-native competitive set
Knowing the set is step one; the payoff comes from closing the recommendation gap—moving your brand from "absent" to "shortlisted" in the prompts that matter. Four moves compound:
- Win the finite shortlist. AI answers name only a few brands per response—often three to five—so displacement, not addition, is the game. Understanding the ceiling on how many brands an AI answer recommends tells you exactly how crowded your target prompts are.
- Design your category. If the model groups you wrong, publish clear, consistent language about the job you do and who you do it for. This is the heart of answer engine optimization—shaping the definition the model repeats back.
- Build off-site agreement. Models recommend brands that independent sources already endorse. Reviews, roundups, and third-party comparisons feed the co-mention clusters that decide your set, so consistent, favorable off-site coverage does much of the heavy lifting.
- Track and re-measure. Recommendation frequency shifts weekly. Treat AI share of voice as a live metric, not an annual audit, so you can prove movement and defend budget.
Done together, these moves are how a low-visibility challenger climbs into its own competitive set—and gets recommended by ChatGPT more often.
Limits and caveats: what this method can't tell you
Be honest about the ceiling: AI answers are probabilistic, so a competitive set is a snapshot, not a ledger. Treat it accordingly.
- Models hallucinate. ChatGPT can invent competitors, misstate features, or fabricate figures—in our own panels we've watched it name vendors that don't exist in the category. Treat any AI competitive pull as a starting point, not a finished report, and validate surprising names against reality.
- Answers drift. The same prompt can return different brands hours apart, which is why single-shot prompting is unreliable and multi-run sampling is non-negotiable.
- Personalization and region skew results. Memory, location, and account history can nudge the set, so document your test conditions.
- Share of voice ≠ market share. A brand can dominate AI answers and still lose deals. The set tells you what buyers see first, not who wins.
These caveats don't weaken the method—they define its correct use. The point of citation and mention tracking is to see the AI-shaped consideration set clearly, then act on it with human judgment. For the wider discipline this sits inside—generative engine optimization (GEO)—Search Engine Land and other SEO trade publications have published solid primers worth reading.
Frequently asked questions
How do I ask ChatGPT who my competitors are?
Ask indirectly. Instead of "who are my competitors," use buyer-style prompts like "what are the best [category] tools for [use case]?" and "[your brand] alternatives?" Run each several times and record every brand named. Direct, self-referential questions bias the answer; recommendation prompts reveal the set buyers actually see.
Why does ChatGPT list competitors I've never heard of?
Because the model clusters brands by how the web talks about them, not by your deal history. A name you don't recognize usually signals a co-mention pattern—shared reviews, roundups, or category language—linking that brand to yours. It's a positioning clue worth investigating, not necessarily a new sales threat.
How often does the AI competitive set change?
Frequently. With live retrieval and model updates, recommendation frequency can shift week to week. Treat your set as a tracked metric with regular sampling rather than a one-time audit; a single snapshot ages quickly and can miss an emerging rival entirely.
Is the AI competitive set the same across ChatGPT, Gemini, and Perplexity?
No. Each engine draws on different sources and retrieval, so the sets overlap but rarely match. A brand can be shortlisted in Perplexity and absent in ChatGPT. Run your prompt panel across every engine your buyers use and compare the sets side by side.
Can I change who ChatGPT thinks my competitors are?
Partly. You can't rewrite the model, but you can influence the inputs—clearer category language, stronger off-site coverage, and consistent third-party comparisons shift the co-mention clusters over time. Measuring before and after is the only way to prove the set actually moved.