Why AI Recommends Your
Competitor Instead of You

A buyer asks ChatGPT which company they should hire, or which software they should buy. The answer names three companies. If yours isn't one of them, you never find out.

Nothing shows up in your analytics. You don't get a lost-ranking alert. The buyer just never calls.

The usual way to check is to type one question into ChatGPT and see what comes back. That tells you very little. AI answers change from day to day, from person to person, and with small changes in wording. A single search can't tell you how often your company gets recommended.

I wrote about checking whether AI cites your site in an earlier post. This one goes a step further. It covers how to measure your share of AI answers at scale, how to find the reasons a competitor gets named instead of you, and what to fix first.

Step 1: Make Sure AI Crawlers Can Read Your Site

Before you measure anything, make sure the engines can get in. Check your robots.txt file for OpenAI's search crawler, OAI-SearchBot, and for PerplexityBot. OpenAI says sites that block OAI-SearchBot won't show up in ChatGPT's search answers.

Then check whether your important content is in the page's HTML or only shows up after JavaScript runs. When Vercel studied AI crawler traffic, it found that none of the major AI crawlers rendered JavaScript. Here's a quick test: open the page, view the source, and search for a sentence from your main content. If it isn't there, AI crawlers probably can't see it either.

Step 2: Measure a Few Hundred Answers, Not One

Tools like Semrush's AI Visibility report track thousands of prompts across ChatGPT, Google AI Overviews, Google AI Mode, and Gemini, and record which companies each answer names.

Compare your domain with four competitors, and pick competitors that each go after a different part of your business. If you only compare yourself with your closest rival, you miss the company quietly winning your second-best service.

Export the prompts, then filter them down to questions a real buyer would ask. Keep questions about providers, software, tools, or "best" and "top" options in your market. Remove questions that name a specific company, and anything about jobs, news, or industry trends.

Write those rules down. Someone on your team, or your agency, will eventually ask how you got your numbers, and "I picked the ones that looked right" won't hold up. Then count the share of answers that name each company.

Before you change anything, run the same export twice, a week or two apart. The difference between those two runs shows you how much the numbers move on their own. If a later change is smaller than that, you can't call it a win yet.

Most tools don't cover every engine. If yours skips Perplexity or Copilot, run your top 20 questions there by hand and log the answers the same way.

Step 3: Split the Results by Service and by Engine

Your overall number will hide the real problem. A company can be named in nearly half of all answers and still barely show up for its most profitable service. Split the answers by service line, and the gap usually shows up right away. That's the example in the scorecard at the top of this page.

Split them by engine too, because each one picks sources differently:

• ChatGPT has leaned heavily on Bing. A 2024 Seer Interactive study found that 87% of its search citations matched Bing's top 20 results. OpenAI now runs its own search crawler too, but Bing is still a good place to start.
• Google AI Overviews breaks a question into smaller questions and cites the pages that answer those. In a March 2026 update, Ahrefs found that only 38% of the pages AI Overviews cited also ranked in Google's top 10 for the same search. Ahrefs changed its method since its earlier study, so the size of the drop is hard to pin down. Either way, ranking well doesn't guarantee a citation.
• Google AI Mode cites a different set of pages. Ahrefs found it cited the same URLs as AI Overviews only 13.7% of the time for the same search.
• Gemini can search Google, but in my tests it often answered buying questions without citing any sources. That means what it already knows about your company carries a lot of weight, and older information can hang around.
• Perplexity runs a fresh web search for every question and shows its sources, so recent, clearly written pages have a better shot there.

One more split is worth doing. "Who are the top providers of X" and "what's the best software for X" can go to completely different competitors. If you sell both a service and a product, check each one on its own.

Step 4: Check Whether Your Own Page Answers the Question

Open the page that should win the topic you're losing. Read it the way an AI engine would, top to bottom, as plain text.

Ask three questions. Does the page have one clear main heading that says what it is and who it's for? Does it mention the specific situations buyers ask about, especially the hard ones? And does it state your size, experience, and results in plain numbers?

Step 5: Find Out Who Wrote the Lists AI Quotes

Search "best [your service]" in Google and Bing, and open the top five lists. Look at who published each one.

You'll often find that a competitor wrote the top-ranking list and put itself first. In the audits I've done, ranked "best of" lists show up again and again as sources for buying questions. That list becomes part of how AI describes your whole category, including what it says your company is weak at.

You can't make a competitor change its own article. You can do three things:

Step 6: Make Your Facts Match Everywhere

Put your homepage, About page, service pages, LinkedIn profile, review profiles, and recent press releases side by side. Compare every number: customers, years in business, size, locations, services.

When those numbers disagree, a list writer has no way to know which one is current. They pick one, and AI engines repeat it. An outdated figure on an old About page can end up in a competitor's comparison table and then in a ChatGPT answer.

The fix is a single approved fact sheet that every page and profile pulls from. It's the cheapest fix in this whole process, and it's usually the first one I'd make.

What I'd Skip

I wouldn't put much weight on one-off screenshots, or on anyone promising a guaranteed "AI ranking." The prompts in these tools are written by the tool to represent common searches, not pulled from your buyers' actual questions. Small topics with 20 or 30 answers will swing from month to month.

I'd also keep expectations for llms.txt modest. No major AI engine has confirmed that it uses the file. It's worth an hour if it gets your team to agree on one set of facts, and not much more than that.

Tie It to Leads, Not Just Mentions

Being named more often is a leading number. Calls, form fills, and booked demos are the ones that matter.

In Google Analytics, set up a report for visits from chatgpt.com, perplexity.ai, gemini.google.com, and copilot.microsoft.com. ChatGPT usually adds utm_source=chatgpt.com to its links, which makes those visits easier to spot.

Know the limits, though. Clicks from Google's AI Overviews and AI Mode show up as regular Google organic traffic, and Analytics can't separate them. Some AI visits arrive with no source at all and land in Direct. Google started rolling out generative AI reports in Search Console in June 2026, but so far they show impressions, not clicks, and only for some sites.

That's why I also add a "How did you hear about us?" question to contact forms. It's the only reliable way to catch the buyer who read about you in ChatGPT and then searched your name. Report leads from AI each month next to your share of AI answers, so you can see whether the work is paying off.

If you can access Bing Webmaster Tools, turn on its AI Performance report. It shows which of your pages Copilot cites and how often.