Why your firm isn't recommended by AI, and the four things that change that

You typed your own specialty into ChatGPT. Maybe "best estate planning attorney in Scottsdale," maybe "fertility clinic with the highest success rates near me." The answer came back in a few seconds: two or three names, a sentence on each, and a confident recommendation.

Yours wasn't one of them.

That moment is why you're reading this, so let's skip the reassurance and get to the mechanics. AI engines don't rank ten blue links; they compose a shortlist. Being outside it isn't a lower position. It's absence. And absence has a cause you can measure and change.

First, how big this is

45% of consumers used an AI tool to find a local business in the past year, up from 6% the year before (BrightLocal, Local Consumer Review Survey 2026). On the other side of that query, ChatGPT recommends just 1.2% of local business locations (SOCi, 2026 Local Visibility Index, 350,000 locations analyzed). Google's AI Overviews now appear in 48.8% of local searches, and AI Mode answers nearly all of them (SteadyDemand, 2026).

What the engines actually cite

Here is the part most firms get wrong. They assume the fix is on their own website. It mostly isn't.

An analysis of more than a million AI citations found that community platforms account for 52.5% of what AI engines cite, and brand domains only 47.5% (Otterly, AI Citations Report 2026). A separate study of citation factors found that 84% of AI citations trace to earned media: things other people published about you, not things you published about yourself (Machine Relations, AI Search Citation Factors 2026).

The strongest correlates with AI visibility in that study weren't keywords or page speed. They were mentions on YouTube (0.737), branded mentions across the web (0.664), and branded anchor text (0.527). Classic backlinks came in at 0.218. Schema markup scored 2.0 out of 10 as a factor.

And the engines disagree with each other. Only 11% of domains are cited by both ChatGPT and Perplexity. Google's AI Overviews show the strongest preference for brand sites (59.8%) while Perplexity leans hardest on Reddit (16.9%). For "near me" questions, Google's AI Mode cites a Google Business Profile 79.8% of the time, while AI Overviews cite the business's own website 73.5% of the time. That is nearly opposite behaviour for the same question (SteadyDemand, 2026).

So "get cited by AI" is not one job. It is five engines, each trusting a different set of sources, and a firm that shows up in one is usually invisible in the others.

The four things the named firms have

Read the firms that do get named for your specialty and the same four things appear every time.

What doesn't work

Writing "for AI." Google's own guide to AI features (updated July 2026) lists what it says you don't need: llms.txt files, "chunking," special AI-friendly formats, or inauthentic mentions. The same guide says what does matter: people-first expert content, a crawlable site, and a maintained Google Business Profile. Meanwhile 73% of websites have technical barriers blocking at least one AI crawler (Otterly, 2026), often a well-meaning security setting nobody has looked at since it was installed.

And guarantees. Nobody can guarantee what an AI system will say. Anyone who promises a position in the answer is selling you something they cannot deliver and, in most licensed professions, something your governing body would not let you repeat.

How you find out where you stand

You measure it the way a client would: the real questions people ask for your specialty in your market, run across all five engines, with the answers recorded word for word. Then you score what came back (who was named, where, and why) and freeze that as a baseline so every claim afterwards can be checked against it.

That is exactly what the free Visibility Audit is. It takes two minutes to request and comes back as a 0–100 score, the actual AI answers, and your three biggest gaps, in a 30-minute readout. One audit per firm. The method is published in full at forevercited.ai/methodology.