Why AI skips your site, and how to find its sources

The reason AI looks past your website

AI tools build answers mostly from what other sites say about a business. In an AirOps study of more than a billion AI citations, third-party pages produced about 85% of brand mentions, with odds of 6.5 to 1 in favor of outside sources over a brand's own domain.

A University of Toronto study pointed the same way, reporting that 91% of answers written by AI cite third-party content rather than brand websites. So the useful question is not why your site was skipped, but which pages the tool read instead.

Why guessing at fixes wastes time

It is tempting to respond by rewriting the homepage or adding new pages. That can help, yet it misses the mark if the tool is forming its view from a directory listing, a review profile, or a news story you have never checked.

A short diagnosis shows where the answer engine's picture of you comes from. Each fix can then go where it will actually change the answer.

Step one: ask what your customers ask

Ask ChatGPT, Perplexity, and Google AI Overviews the questions a customer in your category would ask. Use plain wording, the way a neighbor would ask for a recommendation, and include your town where it fits.

Write down whether your business appears, how it is described, and which other businesses are named. Run the same questions on each platform and compare the replies side by side.

Step two: list the sources behind each answer

When an answer links to or names its sources, record each one. Across several questions, look for the directories, review profiles, and articles that keep showing up.

Then sort what you find into the places AI tools draw on most often:

• Google Business Profile, Yelp, and directories for your industry
• Local media and regional news coverage
• Local business directories and chamber of commerce listings
• Industry publications and the trade press
• Wikipedia and similar high-authority reference sites
• Reddit and other forums or community platforms

Step three: read each source the way the tools do

Read what every source on your list says about you, one page at a time. Note whether it describes your services the way your own website does, and whether anything on it is out of date.

AI systems cross-check what they read, and a mismatch reads as a trust problem. Mark each conflict, each stale entry, and each important source where you do not appear at all.

Step four: give your Google Business Profile a close look

Your profile deserves its own review, because AI Overviews take local data directly from it. Confirm that your hours, photos, services, and description are current.

Check when it was last updated as well. Profiles left untouched for more than 30 days have seen impressions drop.

Step five: measure the gap in mentions

Look at how often your business is mentioned on outside sites, then do the same for the businesses the tools did name. For AI citations, brand mentions matter more than backlinks, and their tie to citation probability is far stronger.

If the named businesses appear in local news, trade blogs, and association directories while you do not, you have found the gap that matters.

Why the diagnosis is worth the effort

The stakes keep rising. AI Overviews appear on about 60% of searches in the United States, and by March 2026 they showed on 14% of shopping queries, 5.6 times the November 2024 level.

Sites that have not adjusted are seeing informational traffic fall by 20 to 40 percent. Meanwhile, a brand cited in an AI Overview collects about 35% more organic clicks than a brand ranking on the same page without the citation.

How AppWT ran this diagnosis on itself

AppWT ran these steps on its own business on October 3, 2026. It asked ChatGPT and Perplexity eight questions that named AppWT and two that did not, then recorded for every answer whether AppWT was named, whether appwt.com was cited, whether twelve checkable facts came back right and whether anything false appeared.

The record showed where to work next: neither engine named AppWT on the question about enterprise work, so publishing that work more clearly became the next fix. The full results are here.

Turning findings into a fix list

When the diagnosis is done, rank each problem by how often its source appeared in the answers you collected. Fix the sources that show up most, then work down the list.

• Wrong or missing information on a frequent source comes first, since they shape every answer that draws on it.
• Missing review profiles come next, because ChatGPT is significantly more likely to cite businesses that keep them.
• Thin outside coverage calls for pitching expert quotes to local news and industry blogs.
• Your own pages should open with a direct answer, since research shows 44% of AI citations come from the top 30% of a page.

Repeat the diagnosis every few months. Refreshing your main pages and your profile on that schedule also tells AI systems your details are current.

Getting help with the diagnosis

AppWT works on both layers, the technical SEO base a site depends on and the GEO work that puts a brand inside AI answers. Depending on what the diagnosis turns up, the next step could be a faster and more crawlable site, content planned for AI visibility, or a review of your off-site presence, scoped to your budget and goals.

Contact AppWT or book a call for a look at how AI search treats your business today.