Why does ChatGPT recommend your competitors? A guide for manufacturers, software companies and online stores

“Who can find out why ChatGPT prefers our competitors?”

Avisible hears this question more and more. A marketing director types the category into ChatGPT, reads the list, and the company is not on it. A competitor is. Sometimes three.

The usual reaction is to assume that AI thinks the competitor is better. In Avisible’s audits, that is often not what the data shows. When AI compares two named companies directly, it often picks the audited company. The problem comes earlier: the company never gets onto the shortlist.

From October 2025 to September 2026, Avisible ran about 90,000 AI queries. This guide explains why competitors get the shortlist and what to do about it. The causes are different for manufacturers, software companies and online stores, so each gets its own section.

In Avisible’s data, these five points explain most of what happens:

1. You often win the comparison but lose the shortlist. When the question names you, AI names you. When the buyer describes a need, the competitor is on the list.
2. Manufacturers: AI recommends by application. Write one page per application and keep the model name wherever the product is sold.
3. Software companies: AI matches a situation to a product. If no source describes you in that situation, you are not in the answer.
4. Online stores: AI gives each brand a lane. Write about everyday use, not only the hero case, and check what marketplaces say about you.
5. The competitor AI names is often not the one you track. Start from the questions buyers ask, not from a list of rivals.

A marine services firm audited by Avisible was chosen 41 times against its main competitor’s 2, in 119 direct comparison answers (76 answers named no winner). An industrial manufacturer was chosen in 77 of 93 comparisons that had a winner. Yet the same marine firm was named in only 7% of open questions in Sweden and 9% in Denmark (base: 90 answers each).

These are two different questions. “Is X better than Y?” starts from your name. “Who are the best providers of …?” starts from the category, and AI builds the list from the sources that describe the category best. Almost every brand Avisible measures is named in 90–100% of answers when the question includes its name. In open questions, without the name, the range is 0–40%. That gap is where competitors win.

So the useful question is not “why does AI like them more?” It is “which questions and which sources put them on the list, and not us?”

For industrial products, AI recommends by application. An industrial equipment maker was named in 3 of 9 answers about asphalt recycling. Those answers named two competitors, and both had dedicated pages for that application. The maker had a machine for the job but no page that said so in plain words.

• Write one page per application. Name the task, the material, the conditions and the model that fits.
• Put the model name everywhere the product is sold. Of 184 answers that named one industrial company, 119 gave no model name. A reseller listing left the model name out of its title, and the product appeared unnamed in 15 Gemini answers.
• Check what resellers say about price. Of 71 buying answers, 67 sent the buyer to a reseller. Only 22 named the maker’s own store. The store price appeared in 0 of 448 answers; a reseller’s lower price appeared in 55.
• Watch your own sister models. The same maker’s sister model was named more often than the target product (67 vs 61 ChatGPT answers). Sometimes the “competitor” is you.
• Check comparison answers for errors. 30% of comparison answers (54 of 180) contained a factual error, against 6% (5 of 89) of plain product questions. In 7 answers AI gave a 1-year warranty instead of 2. A current comparison table on your own site gives AI something correct to repeat.

For software, AI answers “which tool should I use?” by matching a situation to a product: “the best tool for a small team that needs X”, “an alternative to Y for companies in Z”. If no source describes you in that situation, you are not in the answer, even if your product fits.

A technology company audited by Avisible had every technical box ticked: all pages indexable, llms.txt live, all AI crawlers allowed, about 900 words per page. It was still almost never named in open questions. The technology was fine. The words did not describe the situations buyers asked about.

• Name the situation, not only the feature. “For manufacturers with 5–50 sales reps who…” is easier for AI to match than “a powerful, flexible platform”.
• Get into “best tools” lists. AI reads lists and comparison articles. If you are not on them, you are rarely on its list.
• Keep review and comparison site profiles complete. Category, use cases, integrations and pricing model, all current.
• Publish fair comparison pages of your own. Include the cases where a competitor fits better. AI tends to trust pages that do not oversell.

AI gives each brand a lane. A consumer brand won extreme-weather questions in 83–100% of answers but everyday-use questions in only 0–17%. A hotel was described as an energetic city hotel and was named in 0 of 9 answers to a quiet-hotel question. Lanes come from what reviews, forums and your own pages say most often.

• Map your lanes. List the situations where you want to be recommended and check which ones AI already gives you.
• Write about everyday use, not only the hero case. If all your content shows the extreme case, AI will recommend you only for the extreme case.
• Check marketplaces and resellers. Two online stores disagreed on a consumer brand’s shipping terms. AI quoted a US price three times higher than the real one and the wrong shipping terms.
• Look at Reddit and local sources. For the same consumer brand, Reddit was the most-cited source in English. In German it was a German pet shop. Each language has its own sources.

Many companies ask Avisible to compare them with one named rival. The data often shows a different picture. For an industrial equipment maker, the tracked main competitor appeared in 0 of 443 answers, while three untracked brands appeared in about half of them. For an industrial manufacturer, an untracked competitor appeared in 85 answers. If you only watch the rival you know, you miss the one AI actually recommends.

That is why Avisible starts from the questions buyers ask, not from a list of competitors. Whoever AI names in those answers is your competitor in AI search.

Avisible finds out which competitors AI names instead of you, question by question, and why. Avisible’s AI visibility platform shows, for each tracked question, the share of answers that name your brand and the competitor named most often.

For a one-off view, the Avisible Visibility Snapshot (€499, 900 AI queries in ChatGPT, Gemini and Google AI Overviews, delivered in three working days) shows where competitors win and which sources AI reads. The Visibility Audit (200+ buyer questions, six AI systems, four weeks) adds a competitor side-by-side, source mapping and a prioritised 90-day action plan.

Avisible offers both: a one-off Snapshot or Audit, and continuous tracking on the Avisible platform. Use the one-off report to find the gap. Use the tool to see whether it closes over time. Consulting, retained or by project, helps fix what the data shows.

The figures come from Avisible Snapshots and Audits run between October 2025 and September 2026. Each question was asked in several rounds in ChatGPT, Gemini and Google AI Overviews, and in some audits also in Claude, Copilot and Perplexity. Open questions do not contain the brand name. AI answers vary by platform, model version and time; the figures are snapshots of the measured conditions.

• Why is AI recommending our competitors?
• Why doesn't AI search find our company?
• How to improve your AI visibility: start with what AI already reads
• Who can help with AI visibility in Finland?

Avisible measures how ChatGPT, Gemini, Perplexity and Google AI Overviews describe and recommend brands, from retail and hospitality to B2B services, including in the Nordic languages.

The Visibility Snapshot uses 900 AI queries (100 questions, three rounds, three AI systems) to show where AI finds your company, what it says about you and which sources it reads. Delivered in three working days, €499.