How AI assistants decide which business to recommend

Use of AI tools for local business recommendations went from 6% to 45% of consumers in a single year, according to BrightLocal’s 2026 survey. Whatever you think of the exact figure, that is not a niche behaviour any more.

Assistants recommend businesses by reading the open web and, in most cases, a live search index — directories, review platforms, listicles, forum threads and the businesses’ own pages. They do not have a ranking algorithm of their own in the sense search engines do; they synthesise whatever the retrieval step returned.

Which means the lever is the same as it has always been: be present, consistent and well described in the places an assistant is likely to retrieve.

Directory and review platform pages carry disproportionate weight because they are structured and contain exactly the fields the question needs — name, location, rating, category. Listicles and “best X in Y” articles come second, because they answer the question in the shape it was asked.

Your own website matters less than businesses expect, and matters most when it states plainly what you do, where, and for whom. Pages written to rank rather than to inform tend to be skipped, because there is nothing in them to extract.

That this is a new discipline requiring new tactics. It is mostly the old discipline, applied to a reader that cannot infer. An assistant will not work out that you serve the north of the city from a photo of the skyline, and it will not guess your speciality from your tagline.

The second mistake is optimising the website and ignoring the profiles. If an assistant cites a review platform and your profile there is thin, the website cannot compensate.

A consistent name, address and category everywhere. A healthy review profile with recent reviews, because rating and count are the two fields most likely to be quoted. Plain descriptions of what you do in the words a customer would use. And presence in the listicles and directories for your category, which is unglamorous and effective.

Structured data helps machines parse the page, though it is not the deciding factor. Being genuinely the answer to the question is.

BrightLocal’s Local Consumer Review Survey 2026 puts AI tool usage for local recommendations at 45%, up from 6% in 2025, while Google’s share of review reading fell from 83% to 71%. One survey, one market, early days — but the direction is consistent with what everyone is seeing in their own traffic.

Worth pairing with the other finding from the same survey: the average consumer now consults six review platforms before deciding. Breadth of presence matters more than it did.

Ask the assistants directly. Run the queries a customer would run, note whether you appear and which sources get cited, and repeat monthly. It is crude and it is more than most businesses do.