Why does ChatGPT recommend competitors instead of your brand? AI visibility monitoring makes the gap visible: other companies appear consistently, while your brand is absent or placed at the end of the shortlist.
Instead of concluding that "AI is wrong," run a competitor analysis for GEO: identify the queries that trigger recommendations, the sources supporting each competitor, the ChatGPT Share of Voice gap, and the evidence your brand lacks.
The first rule: one measurement is a hypothesis, not a diagnosis
AI models are stochastic. The same query in ChatGPT can give different answers on different days, name three competitors in one answer, and two in another. Therefore, before drawing conclusions:
- Make 3-5 repetitions of the same query on different days;
- compare with the answers of other models (Gemini, Claude, Perplexity);
- record the date, mode (with/without web search) and model;
- see if competitors repeat themselves from answer to answer.
Only after that there is material that you can work with. One screenshot is not a reason to change the content strategy.
Follow the chain: query → answer → source → competitor
If you look only at the name of the competitor, the conclusion will almost always be superficial. It is more useful to lay out the entire chain and see exactly where the break occurs.
| Layer | What are we looking for | An example of an insight |
|---|---|---|
| Query | Which user intent triggered | "This is a comparative query — the model is looking for not one player, but a selection" |
| Answer | What wording does the model use | "A competitor described as a 'leader for B2B SaaS' is a ready-made association" |
| Sources | What the model relies on | "An AIN article is cited in which we are not and two directories where we are passive" |
| Competitor | Who exactly appears and why | "Competitor consistently in 7 out of 10 queries is the leader of the reference group" |
Without the first three layers, the conclusion "we need to rewrite the landing page" is almost always wrong.
Types of Competitors in AI Responses
Not all competitors get into the answers for one reason. It is useful to immediately classify who exactly you are dealing with:
| Type | How does it appear in the answer | What does it mean for you |
|---|---|---|
| Category leader | Appears in almost every category query | Stable association with the market — you will have to bypass due to niche |
| Expert Brand | Appears in narrow/professional queries | Strong content and PR in specialized media |
| Brand with a strong product | Appears in feature comparisons | Strong product pages, documentation, and FAQs |
| "Catalog" competitor | Appears because it is well represented in ratings and selections | A PR channel and presence in the lists of |
| Random competitor | Appears unstable | Most likely noise, do not concentrate the resource |
The first four types are different tasks. The leader of the category "overtake in content" is very expensive; It is better to work with it through niche and your own reference group.
The same pattern appears in our study of how ChatGPT recommends smartphone brands in Ukraine: the model does not simply name a leader; it groups brands by use case. That context matters when diagnosing why your brand is absent.
Four levels of answer parsing
When it is clear that the competitor appears steadily, then we break down the answer into four levels.
1. What type of competitor does the model name?
- who exactly appears;
- whether it is one brand or several different ones;
- whether the brand appears across different query types (comparison, category-level, problem-oriented, or all of them);
- is it a Ukrainian player or a model pulling a global leader into a local context.
2. Why exactly does the model recommend it
We record the exact formulation of the model. Not "AI praises a competitor", but specifically: "has experience in fintech", "convenient for a team of up to 10 people", "often recommended for speed of implementation". These phrases are a ready-made artifact for further work: you need the same clear context for your brand.
3. What sources does the model rely on?
If the model shows sources, this is the most valuable layer. Let's see:
- whether the competitor's website is cited (then the task is in its own content);
- whether editorial materials (then the task is PR);
- whether ratings and catalogs (then the task is to be present in the lists);
- whether reviews and platforms like G2/Clutch (then the task is to work with reviews and cases).
If you want to go deeper, see separately the material How to analyze sources based on AI.
4. What you're missing in the same context
The most important level. The question is not "what is strong in the competitor", but "what we do not have in the same field". Possible findings:
- there are no materials for this scenario of choice;
- the brand does not appear in external sources cited by the model;
- the site does not explain your role in the category in one sentence;
- competitors have comparison pages, you don't;
- the model does not "see" an understandable advantage.
Six Reasons Why AI Chooses a Competitor
From what is repeated in practice:
- The competitor has a clearer explanation of "for whom". AI works well with clear associations. If it is easy to understand about the brand for whom it is and what task it closes, the model often includes it in response.
- More external confirmations. Mentions in industry media, ratings, reviews, forums — each of them gives the model a reason to trust.
- Stronger page formats for the intent. Comparisons, FAQs, guides, and case studies are easier to cite. A site without them loses visibility in answers that use web search.
- Stable context in training data. If a brand consistently appeared on the open Internet 1-3 years ago in the same context, the model "remembers" this association even without web search.
- Technical accessibility of the site. If search crawlers such as
OAI-SearchBot,Claude-SearchBot, orPerplexityBotcannot access the site, its pages are less likely to appear in search-backed answers. Training controls such asGPTBotandGoogle-Extendedserve a different purpose. - The brand name competes with another meaning. If the brand name is homonymous to a common term or another brand, the model is confused, and a more "understandable" player is in the answer.
Usually, a combination of 2-3 reasons works in a specific answer, not one. Therefore, the action plan is also always combined.
Competitive map: what should be clear from it
Instead of a list of "who bypasses us", it is useful to build a map that shows the reference group and dynamics.
The minimum that should be in the card:
- 5-10 competitors who consistently appear nearby;
- the share of mentions of each in your query pool;
- which query types they are strong in (category-level, comparison, or problem-oriented);
- what sources "carry" them (their website / media / catalogs);
- changes between measurement periods.
The card does two important things: it reduces anxiety ("we are not alone — this is the structure of the market") and focuses actions ("if objectively we are now in the middle group, we are fighting to get into the top 3, not to bypass the world leader").
How to translate a gap into an action plan
Specific hypotheses depend on what the previous analysis has shown. Working matrix:
| What do we see in the answer | Hypothesis to test |
|---|---|
| Competitors are strong in external media | PR campaign: expert columns, interviews, and comments for journalists |
| Enter through comparison pages | Their comparisons in a stronger format (tables, criteria, honest pros and cons) |
| Clearly described role in category | Positioning work: one sentence "who are you for + what are the advantages" on the home page and service pages |
| Cited through directories and ratings | Apply to relevant ratings, update profiles, and collect reviews |
| The competitor stays on one type of request | We focus the content on this type, without dispersing |
| Competitor appears in the response without web search | This is a signal of the training layer — a long cycle of PR and stable mentions |
Each hypothesis should lead to a specific action and an expected effect, tested in the next measurement.
Common errors in interpretation
- "AI is wrong" as a conclusion. AI is a mirror of what is visible about you and your competitor on the open Internet. There is no point in arguing with the mirror.
- Bet on one answer. Without repetition, the conclusion is fragile.
- Fighting the leader of the category "head-on". Usually expensive and slow. The fastest way is to niche and reach the top of the reference subgroup.
- Ignoring query context. A competitor may appear in a category query and a comparison query for different reasons, so each gap requires a different response.
- Verification in only one model. ChatGPT, Gemini and Claude can show very different pictures. The real one is at the intersection.
Frequently Asked Questions
Does the appearance of competitors mean that we are "bad"? Not necessarily. In most niches, AI always names 2-5 competitors — this is a normal response structure. The question is not "are we not there", but "are we on the list and in what context".
Is it possible to "ask" a model to recommend your brand more often? Not directly. Models rely on signals from public sources, so the work focuses on improving those signals.
How quickly do you see the effect after actions? There is no guaranteed timeline. Search-backed answers may change after recrawling and source updates, while responses without live web search can take much longer to reflect new information.
Is it worth analyzing every competitor that appears? No. Focus on recurring competitors that appear in at least three prompts from the pool. Treat isolated mentions as background until they repeat.
What to do if a brand that is not a competitor to us appears in the response? This is also a signal. Perhaps the model has its own categorization logic, and the user considers you along with this brand at the time of choosing. It is worth adding it to the monitoring and seeing the dynamics.
What else to read
- How to Analyze AI-Based Sources
- Which pages of the site are most often included in AI responses
- How to Turn AI visibility report into a plan for SEO, content, and PR
- What Is AI Visibility And Why Businesses Are Not Enough Anymore SEO
How we do it in VYDAI
You can analyze one or two competitors manually. At 5-10 competitors, four models, 50-80 prompts, and repeated checks on different days, each measurement already contains hundreds of answers.
VYDAI automates this analysis. For each prompt, you can see which competitors appear beside your brand, how share of voice changes between measurement periods, how models describe competitors, and which sources they cite. The resulting competitive map shows the reference group and its movement over time.
If you want to see how it looks on your competitors, you can create an account or see demo. How exactly to react to the found gap is up to you; We will be there and show the logic of decisions.
