Why this feels so unfair
You know the competitor. You have seen their work. Being told an assistant prefers them reads like a verdict on your business, and it is not one. A model has no way to assess whether their panel upgrades are any good. It can only read what the internet says, count how consistently it says it, and repeat the names that come up most.
So the useful question is not "why them". It is "what does the internet say about them that it does not say about me". That is answerable in an afternoon, and it is nearly always a short list.
The three gaps that produce this exact symptom
- Roundup absence. Somebody, often a local publication or a lead-generation site, published a "best [trade] in [city]" page. Your competitor is on it. You are not. That single page can carry an assistant's whole answer.
- Job-level language. Their site or their reviews use the words a customer types ("sewer line replacement", "same-day AC repair") and yours say "quality service since 1998". Assistants match on the specific, not the reputable.
- Entity thinness. Their name, address, phone, service area, and category agree across every source. Yours differ in two places, so the assistant is less certain the business exists as described and quietly prefers the safer name.
What to do with the audit below
Pick the one competitor the assistants name most often, not the one you dislike most. Fill in the table honestly. Every row where they have something and you do not is a concrete task, and the rows are roughly in order of how quickly they pay off.
Do not try to close all of them. Closing the top three rows moves more than doing all ten badly, and the top three are usually a profile category, one roundup, and ten reviews that mention the actual job.
Competitor coverage audit
Search each row for your business and for the competitor the assistants name most. Mark yes or no. Print it, or copy it into a spreadsheet.
| Source to check | How to check it in one minute | Why an assistant cares |
|---|---|---|
| Google Business Profile completeness | Search both business names on Google. Compare categories, services listed, hours, photos, description. | Nearly every local source an assistant reads is derived from this profile. |
| Primary category match | Whose primary category matches the job customers actually ask for? | A wrong primary category mislabels you across every downstream source at once. |
| Best-of roundups | Search "best [trade] in [city]" and open the first five non-map results. Who is listed? | These pages are written in exactly the shape of the question, so they are retrieved constantly. |
| Review count and recency | Compare total reviews and the date of the most recent one. | Stale review flow reads as a business that may no longer be active. |
| Review text specificity | Read the last ten of each. Count how many name a specific job or product. | Assistants quote the specific. A five-star review saying "great!" contains nothing to quote. |
| Service pages on your own site | Does each service have its own page with the service name in the heading? | One page per service gives the assistant a retrievable answer to a specific question. |
| Service area stated in words | Is the list of towns you serve written as text, not only implied by a map widget? | Text is readable. A map embed is not. |
| Name, address, phone consistency | Check the profile, the site footer, and two directories. Do all four agree exactly? | Disagreement makes an assistant less certain which entity you are. |
| Local press or association mentions | Search the business name plus the city, excluding your own domain. | A third party saying it counts for more than you saying it. |
| Anything with a date on it | Is there a page on your site published in the last ninety days? | Recency is one of the few signals a retrieval system can check cheaply. |
Do this against one competitor, not five. The overlap between competitors is high, and one honest comparison produces a shorter, truer task list than five rushed ones.
You are looking for the term
Share of voice
The pattern you are describing has a metric behind it: share of voice, the proportion of AI answers in your category that name you rather than a competitor. It is the number most agencies mean when they talk about competitive AI visibility, and it is why the fix is measured in coverage rather than in effort.
- Can I get a competitor removed from an AI answer?
- Not unless the answer states something factually false about your business, which you fix by correcting the underlying source. Assistants are not a listing you can dispute your way up.
- How many roundups do I need to be on?
- Fewer than you would guess. Two or three pages that genuinely rank for "best [trade] in [city]" change what gets retrieved far more than twenty low-quality directory listings.
- 01Tips to improve your local ranking on GoogleGoogle Business Profile Help
- 02Guidelines for representing your business on GoogleGoogle Business Profile Help
- 03Top ways to ensure your content performs well in Google's AI experiencesGoogle Search Central
- 04GEO: Generative Engine OptimizationAggarwal et al., arXiv
Every link above was opened and confirmed reachable on September 1, 2026.
The automated version
See what ChatGPT, Claude, Gemini and Perplexity say about your business
The free scan runs the buyer questions for your trade and city across all four assistants, shows who gets named instead of you, and lists the sources those answers came from. No signup, about thirty seconds.
Run my free scan