Why is my competitor's brand mentioned in AI answers but mine isn't? How we diagnose it
Why is my competitor's brand mentioned in AI answers but mine isn't? The diagnostic order we follow on client programs, with the Promptwatch reports we open.
This is the most common first-call question we hear. A founder or marketing lead has asked ChatGPT about their category, and a competitor is named while they are not. Sometimes the competitor is smaller, newer or worse. The founder wants to know what the competitor did.
The honest answer starts with "it depends", then moves quickly to a method. We run client programs on Promptwatch, and the diagnosis below is the order we follow.
First, confirm it is real
Before explaining anything, we check whether the gap holds up. AI answers vary from run to run, and a single manual check is a single sample. So we build a prompt panel around the questions that matter and run it daily per model, country and language. Then we look at Share of Voice and Position in Answer against the named competitor.
Sometimes the gap is real across engines. Sometimes it exists on one engine and not the others, which changes the diagnosis. Sometimes it shrinks to nothing, and the founder was reacting to one unlucky answer. We would rather find that out in week one than after a plan.
Promptwatch's competitor heatmap helps here, because it compares any two dimensions, such as brands against models or topics against countries, and shows where the gap lives.
Second, check access
If AI crawlers cannot read your pages, nothing else matters. We read robots.txt for rules that block GPTBot, OAI-SearchBot, ClaudeBot, PerplexityBot or Google-Agent. We check that the content is present in the HTML the bots receive and not hidden behind scripts. Then we open Agent Analytics, which shows IP-verified crawler logs with errors and failure explanations.
The free Agentic Readiness Check gives a quick first pass if the client has not yet connected a log source. Crawler logs start on Promptwatch's Professional plan, so we say that early in scoping.
A TreeHouse case study published by Promptwatch, covering Pararius and Huurwoningen.nl, describes AI search growing to about 25% of organic conversions after crawler access was fixed. We cite it as their experience, and we use it to remind clients that access comes first.
Third, compare the sources
When a competitor is named and the client is not, the engine usually relied on specific pages. We open the citation reports and ask: what are they citing for the competitor, and what are they citing for us?
Three patterns show up often:
- A third-party page names the competitor and not us. A comparison post, a review roundup or a forum thread. The fix is outreach, a correction or a contribution, not a new page on our site.
- The competitor has a page that answers the sub-questions directly, and we do not. The fix is content.
- Reddit, YouTube or LinkedIn carries the discussion. The fix is participation, done honestly.
Promptwatch reports citations at page and domain level. Page Tracker follows the specific URLs we care about, ours and the competitor's. The offsite mentions report shows where a brand is named on a third-party page without a link, which is often the cheapest win.
Fourth, check coverage
We look at query fan-outs for the prompts we lose, then compare them with what the client's site covers. Promptwatch's content gap analysis gives coverage scores, and the Answer Gap Report shows missing answers. If there is no honest page that answers a sub-question, we do not try to trick the engine. We write one.
Fifth, read the answers themselves
Sometimes the engine mentions the client but gets facts wrong, or describes them thinly. We read full responses for brand prompts, check the sentiment score for each, and look at the weekly AI summary. The Brand Book and brand aliases decide what counts as a mention, so we review those to make sure variants of the name are recognized.
What we do not promise
We do not promise the client will be named after our work. We cannot see inside the models, and they change. What we can promise is a clear diagnosis, a prioritized list and a trend to watch. We also tell clients not to compare themselves with the competitor's single best answer. Share across many runs is what matters.
How we pick the first fix
We pick the cheapest fix that the evidence supports. If crawlers are blocked, we unblock them. If a single third-party source dominates, we go after that. If the content is missing, we publish. Promptwatch's Action Board helps with the order, since it gives a prioritized GEO to-do list with evidence attached, and its Discovery agent suggests up to five actions a week.
We change one thing per topic, record the date, and leave it alone for a few weeks.
If you are doing this yourself
Start with a stable prompt set, check access, then compare sources before touching content. Promptwatch's free Explore plan gives you 10 prompts on ChatGPT, enough to see how the competitor comparison looks. The paid plans open all models, and Professional adds crawler logs for the access step.