Why ChatGPT recommends your competitors instead of you: 7 reasons
ChatGPT recommends businesses it can find, read and confirm across several sources. If it names your competitors, the usual causes are a site that blocks AI crawlers, weak presence in Bing's index, no written answers to customer questions, or too few mentions of you outside your own website. Below are 7 reasons and what to do about each.
Reason 1. Your site blocks AI crawlers
AI engines read websites through their own bots. OpenAI runs three: GPTBot collects data for training, OAI-SearchBot indexes pages for ChatGPT search, and ChatGPT-User opens a page when a user asks for it in a chat. There are also ClaudeBot, PerplexityBot and Google-Extended.
If robots.txt disallows all bots, or these ones specifically, the AI engine can't see your current content. The second common case is bot protection at the hosting or CDN level that blocks these crawlers before they ever reach robots.txt. Site owners usually have no idea it's happening.
What to do: open your-site/robots.txt and check for Disallow rules for these user agents. At a minimum, allow the search bots: OAI-SearchBot, ChatGPT-User and PerplexityBot. You can block GPTBot separately if you don't want your content used for training; according to OpenAI itself, these settings are independent. Then review your host's bot protection rules.
Reason 2. Your site isn't in Bing's index
ChatGPT with search enabled relies to a large extent on Bing's index. A site can rank decently in Google and barely exist in Bing: nobody added it, no sitemap was submitted, and half the pages aren't indexed.
The same logic applies to Gemini and AI Overviews, except they lean on Google. If a page isn't indexed, a search-enabled AI engine won't find it, however good it is.
What to do: add your site to Bing Webmaster Tools, submit sitemap.xml and check how many pages are indexed. Do the same in Google Search Console. Your sitemap should load and list only live pages, with no redirects or errors. Bing supports the IndexNow protocol, which lets you notify it about new pages right away.
Reason 3. Your site doesn't answer customer questions
People ask ChatGPT specific things: "how much does AC installation cost in Warsaw", "which dental clinic in Kyiv is open on Sunday". The AI engine builds its answer from pages where this is written out in text. If you have one "Services" page with a list of names and "call for pricing", a competitor with detailed pages has an easier path into the answer.
The second problem is key text hidden in images, PDF price lists, sliders or loaded by scripts. Not every crawler executes JavaScript, so what a person sees in the browser may never reach the bot.
What to do: create a separate page for each key service. Put a direct answer in the first paragraph: what it is, who it's for, starting price, timing, location. Add pages for decision-stage questions such as "how to choose", "how much it costs" and comparisons of options. Keep key text in HTML. The full process is in our guide on how to get into ChatGPT answers.
Want to know where your website stands today? Free 60-second AI visibility audit in Telegram.
Check my siteReason 4. No structured data or proper meta tags
Structured data lets a machine understand a page without guessing: here's the business, its address, opening hours, services and prices, and here's a block of questions and answers. Many competitors have Schema.org markup in JSON-LD, and many small businesses don't.
Title and meta description belong here too. When every page has the same title, like "Home" or just the company name, it's harder for search engines and AI to work out what each page is about.
What to do: add JSON-LD using the Schema.org vocabulary: Organization or LocalBusiness on the homepage, Service on service pages, FAQPage wherever you have questions and answers. Write a unique title and description for every page. You can add an llms.txt file as an extra, but to be honest: it's a format proposed in 2024, it hasn't become an official standard, and support by the major AI engines isn't guaranteed. More in our guide on what llms.txt is.
Reason 5. The AI isn't sure who you are
A model links mentions of a company into a single entity: name, address, phone, website, social profiles, industry. If your site shows one name, Google Maps another, a directory an old address and Instagram a third phone number, the AI may not realize they're the same business. A competitor with consistent data everywhere is easier to trust.
Sometimes it's simpler: the site never states plainly what the company does, where it operates and who it serves. A person gets that from the photos. A model doesn't.
What to do: make your name, address and phone (NAP) identical on your site, Google Business Profile, directories and social media. Create an About page with facts: founding year, cities, services, team. In your Organization markup, list official profiles using the sameAs property.
Reason 6. Few people write about you outside your site
For questions like "who's the best…" or "who would you recommend…", the AI relies on external sources: reviews, rankings, directories, industry media and discussions. If a competitor appears in five roundups and you appear in none, your own website won't outweigh that.
We see this in our own case. After we launched a GEO setup on insync.com.ua, citations of the site in ChatGPT went from 0 to 24 in about a week and a half, and for 4 out of 4 branded queries the AI knows the agency and restates its positioning. For category queries ("best agency for…"), InSync isn't named yet. The website covers branded questions; category questions are decided by external mentions.
What to do: list the sites AI engines already cite on your topic. Perplexity is a handy starting point because it shows the sources behind each answer. Get listed in industry directories and rankings, offer expert comments, publish case studies on external platforms and arrange mutual mentions with partners.
Reason 7. Few fresh reviews and a weak maps profile
For local queries like "dentist near me" or "brunch café downtown", map listings and reviews matter. Gemini and AI Overviews rely on Google data, so a complete Google Business Profile and active reviews affect who they show.
A weak profile looks like this: empty categories, no opening hours, old photos, the last review a year ago, no owner replies. The competitor's profile is complete and up to date, so they make the list.
What to do: fill in your profile completely: categories, services, hours, photos, description. Ask for a review after every successful job and reply to all reviews. Keep prices, offers and schedules current on your site so the data matches everywhere.
Where to start: the order of checks
Go from simple to hard: first whatever blocks access, then content, then external signals. Technical issues are usually fixed within a few days, while mentions and reviews build up over months.
| Reason | How to check | What to do |
|---|---|---|
| Crawlers blocked | robots.txt and your host's bot protection | Allow OAI-SearchBot, ChatGPT-User, PerplexityBot |
| Not in Bing's index | Bing Webmaster Tools | Submit sitemap.xml, set up IndexNow |
| No answers to questions | Ask ChatGPT your customers' questions | Service pages with the answer in the first paragraph |
| No markup | Schema.org validator | JSON-LD: Organization, Service, FAQPage |
| Unclear entity | Compare NAP on your site, maps and directories | Make the data consistent |
| Few mentions | Answer sources in Perplexity | Directories, rankings, publications |
| Weak reviews | Your Google Maps profile | Complete the profile, collect reviews |
A quick self-check is described in our guide on how to check whether ChatGPT sees your business. We've automated the technical part and the regular answer pages in GEO Autopilot: a free 8-factor audit in 60 seconds in a Telegram bot, then setup, 4 new pages a month and a monthly before-and-after measurement. Mentions in AI answers usually appear within 2–4 months.
FAQ
Why does ChatGPT name different companies for the same question?
AI answers aren't fixed. They depend on the wording, whether search is on, location and chat history. That's why visibility should be checked on a set of several queries and measured repeatedly; one check in one chat tells you very little.
If we block GPTBot, will ChatGPT stop recommending us?
According to OpenAI, GPTBot collects training data, while OAI-SearchBot and ChatGPT-User handle ChatGPT search. They're configured separately in robots.txt. You can opt out of training and keep your site available to search.
How long until ChatGPT starts mentioning us?
Technical fixes take days, but mentions in answers usually appear within 2–4 months. On our own site, ChatGPT citations went from 0 to 24 in about a week and a half, yet we're still not named for category queries. Branded and category queries grow at different speeds.
Will an llms.txt file alone help?
Unlikely on its own. It's a format proposed in 2024 that hasn't become an official standard, and the major AI engines aren't obliged to use it. It's easy to add, but crawler access, indexing, content and markup come first.
Can anyone guarantee the top spot in a ChatGPT answer?
No. AI answers change and draw on many sources, so nobody can guarantee a "#1 in ChatGPT". Honest work looks like this: fix the causes, measure mentions regularly and show the before-and-after trend.