You ask ChatGPT for the best options in your category. Your main competitor is named first, with a friendly sentence about why. You aren't named at all. You try Gemini and Perplexity and it's the same story.

There's usually a findable reason, and usually more than one. Here are the seven we see most, roughly in the order you should check them.

Why it matters

For many buyers, this answer is the shortlist. In a March 2026 G2 survey of 1,076 B2B software buyers, 51% said they now start research in an AI chatbot more often than in Google, up from 29% eleven months earlier. 69% had picked a different vendor from the one they planned to because of what a chatbot told them, and a third bought from a vendor they'd never heard of before. G2 sells software reviews, so it has an interest in this story, but the direction matches every other study we've read.

1. Assistants can't read your site

Each AI company runs separate crawlers for separate jobs. Some collect pages to train models. Others fetch pages to answer questions and cite them. Block the second kind and you can vanish from that assistant's answers, however good your pages are.

  • OpenAI says plainly that "sites that are opted out of OAI-SearchBot will not be shown in ChatGPT search answers." GPTBot is its training crawler, and blocking it doesn't affect search.
  • Anthropic says blocking Claude-SearchBot or Claude-User "may reduce your site's visibility" in Claude's answers. ClaudeBot is for training.
  • Perplexity uses PerplexityBot to build the index it answers from, and Perplexity-User to fetch pages while answering.
  • Google's AI Overviews and AI Mode use the normal Google index. If Googlebot can crawl and index a page, it's eligible. Google-Extended doesn't affect them, but it does cover grounding in the Gemini app, so blocking it can cost you Gemini answers.

How to check. Open yoursite.com/robots.txt and look for any of those names under a Disallow rule, or a blanket block on all bots. Then ask whoever runs your hosting, CDN or firewall whether AI crawlers are blocked or challenged. Some security services have a one-click setting for this, and it's easy to switch on without realizing it covers search as well as training. Your server or CDN logs show whether the real crawlers are getting pages or errors.

How to fix. Allow the search and answer crawlers. Whether to allow training crawlers is a separate business decision, and you can block those without disappearing from answers.

2. Your pages don't answer the question

When an assistant fans a question out into smaller searches, it looks for facts: what the product is, who it's for, what it costs, where it works, what it connects to and how it compares. Plenty of B2B sites answer none of these in plain words. The homepage says "the operating system for modern teams", pricing says "contact us" and the comparison page doesn't exist.

The assistant can't recommend you for "best project management tool for a 30-person architecture firm" if nothing it can read connects you to architecture firms or teams of that size. Your competitor's site has a page called "Project management for architects", so it gets named.

How to check. Read your homepage, pricing page and main product page as if you'd never heard of the company. Can you tell, in one read, who it's for and roughly what it costs? Then view the page source, or turn JavaScript off, and see whether that text is still there.

How to fix. Say the plain facts on normal HTML pages: the category you're in, the kinds of companies you serve, starting prices or at least a range, the countries you work in, your main integrations. Add pages for the use cases and comparisons your buyers search for. Google's own guidance is that you don't need special files or a special writing style for AI. You need clear, useful pages that crawlers can read.

3. You're missing from the cited pages

This is the most common cause, and the most fixable. For any buyer question there's a small set of pages the assistants keep citing: a couple of "best X for Y" roundups, a G2 or Capterra category page, a comparison article, maybe a directory. If your competitor is on those pages and you aren't, the assistant is only repeating what it read.

How to check. Run your five most important buyer questions in ChatGPT, Perplexity and Google, with web search on. Open every cited source. For each one, note whether it mentions you, your competitors, or both. Within an hour you'll have a list of pages that recommend your competitor and not you.

How to fix. Work that list. Ask roundup editors to consider you when they next update the article, and give them something worth writing about. Complete your review-site profiles and ask happy customers for reviews that mention specifics. Get listed in the directories that turn up. Being cited pays off in clicks too. Seer Interactive found brands cited in AI Overviews got 35% more organic clicks and 91% more paid clicks than brands that weren't.

4. Nobody's discussing you

Assistants lean heavily on what people say to each other. In Peec AI's analysis of 30 million sources cited by five AI platforms, Reddit was the first or second most cited site on every one. YouTube was the single most cited domain in Google's AI Overviews in an Ahrefs study from early 2026. If the thread "What do you use for X?" has forty replies recommending your competitor and none about you, that's the consensus the assistant picks up.

How to check. Search Reddit and YouTube for your category and your competitors' names. Which threads and videos keep appearing in your citation list from step 3?

How to fix. Take part honestly. Answer questions in your category's subreddits and forums as yourself, say where you work, and be useful even when the answer isn't your product. Communities spot and punish astroturfing fast, and it backfires. Make videos that answer buyers' questions: walkthroughs, honest comparisons, setup guides. Encourage customers who like you to say so where they already talk.

5. The web describes you wrongly

Sometimes you are mentioned, but as the wrong thing. The assistant thinks you only serve enterprises, quotes a price from 2023, lists a product you retired, or mixes you up with a company with a similar name. It won't recommend you for a question you don't seem to fit.

How to check. Ask each assistant directly: "What is [brand]?", "Who is [brand] for?", "How much does [brand] cost?" and "What do people complain about with [brand]?" Note every error and look at the sources it cites for each.

How to fix. Correct the facts where they live. Update your LinkedIn, Crunchbase, review-site and directory profiles so they describe you the same way your site does. Ask publishers to correct outdated articles. If you share a name with another company, make the difference obvious: category, location and product names, used consistently. Don't edit your own Wikipedia article. Its conflict-of-interest rules strongly discourage it. Suggest changes on the article's talk page instead.

6. You're strong in another country

A UK firm that's well known in London can be invisible to a buyer asking from New York, and the reverse. Assistants search from or for the buyer's location. OpenAI notes ChatGPT can use an approximate location from the user's IP address. The answer then draws on local review sites, local press and local directories.

How to check. Make sure you're checking from your buyers' country. A VPN works for a spot check. And make sure the questions say what a local buyer would say: pounds not dollars, local terms, local regulations.

How to fix. Earn local sources: local review platforms, trade bodies, regional press and partner listings. Give each market a page with local pricing, currency and proof.

7. Your competitor is simply better known

Sometimes there's no single gap. Your competitor has more reviews, more press, more threads and more videos, and the assistant reflects that. Ahrefs' study of 75,000 brands found web mentions correlated with AI Overview visibility at 0.664, far ahead of backlinks at 0.218. Brands in the top quarter for mentions averaged about ten times as many AI Overview mentions as the quarter below them, and the bottom half barely showed up.

How to fix. Treat it as a long campaign, not a technical fix. Aim at specific questions rather than "more mentions" in general. Pick the five buyer questions you most want to win, find the sources behind them, and put your PR, content and review effort there. Google warns that "seeking inauthentic 'mentions' across the web isn't as helpful as it might seem", and its spam systems are built to ignore them.

Where to start

Put the checks in order of how quickly they pay off.

  1. Access, in minutes. robots.txt and firewall. If you're blocked, nothing else matters.
  2. What assistants say about you, in an hour. Ask the four questions from cause 5 on each assistant and fix the wrong facts at their source.
  3. Your own pages, in a week or two. Plain facts, use-case pages, comparison pages.
  4. The cited pages you're missing from, over a month or two. This is where most of the movement comes from.
  5. Being discussed more, from then on. Reviews, community, video, press.

Then measure over time, across many answers, because any single answer can mislead. Our GEO audit guide turns all of this into a day's work with a spreadsheet, including how many answers you need before a change is real. Our report on what AI answers cite shows which sources matter most across the assistants.