Short answer
You cannot control what AI assistants say, but you can raise the odds of being named. Be easy to find and read on the open web, state clearly who you are and what you offer, publish specific answers backed by real data, and be mentioned by other trustworthy sites. Assistants draw on live search results and on what they learned from the web. No tactic guarantees a mention.
What counts as being mentioned by an AI assistant?
There are three levels. A mention is your brand name in the answer text. A citation is a link to one of your pages in the sources. A recommendation is the assistant steering the reader to you or naming you among the best options. A mention with a citation is the strongest combination, because the reader can click through and see the claim is sourced.
To see where you stand before changing anything, run a small fixed set of customer questions in several assistants, logged out, and repeat each one, because answers vary from run to run. What AI visibility is and how to check it gives the full method.
Why do assistants name some brands and not others?
An assistant builds an answer from two places. One is what a model absorbed in training, which gives it a sense of which brands exist and what they are known for. The other is live web search, where it fetches pages and quotes them. Vendors do not publish exactly how they choose, so anyone who sells you a formula is guessing.
Two findings help. Ahrefs' analysis of 1.9 million AI Overview citations found that about 76% of cited pages ranked in Google's top ten, so ordinary search visibility is the entry ticket. And the 2024 Princeton and IIT Delhi paper on generative engine optimization found that adding statistics and cited sources raised visibility in generated answers by up to roughly a third, while keyword stuffing was the weakest of the nine tactics tested.
Practitioners also report that brands named across many independent sites get named more often. That is plausible, but it is measured less carefully than the two findings above.
What can I do to get mentioned?
- Say who you are, the same way everywhere
Use one business name and a one-sentence description of what you do, for whom and where, on your homepage, About page, Google Business Profile and other profiles. Consistent facts are easier to repeat correctly.
- Answer real questions directly
Give each page a question heading and a short direct answer first, then the detail. SE Ranking's data suggests sections of roughly 120 to 180 words between headings are cited most.
- Add data only you have
Original numbers, first-hand examples, named sources and dates give an assistant something specific to quote. In SE Ranking's study, pages with many concrete data points were cited more than data-light pages.
- Let the crawlers in
Make sure robots.txt and your host's bot protection allow the crawlers behind AI search, and that key text is in the page's HTML and not only added by scripts. See which AI crawlers to allow or block.
- Be present beyond your own site
Earn real mentions where customers and assistants look: industry directories, review sites, local press, partner pages and honest answers in communities. Ask happy customers for genuine reviews. Fake reviews and paid placements tend to backfire.
- Keep pages fresh and attributed
Show the author and a last-updated date, refresh old facts and fix pages that contradict each other.
What does not work, or is unproven?
- Keyword stuffing. It was the worst performer in the Princeton and IIT Delhi test.
- llms.txt. Google says it does not use it, and no major AI answer engine has confirmed reading it. Does llms.txt work explains.
- Schema markup as a citation trick. Ahrefs' May 2026 study of 1,885 pages that added JSON-LD found no meaningful lift in AI citations, though every page already had 100 or more AI Overview citations. Schema still helps rich results and makes entities clearer.
- Hidden text aimed at AI tools. It is manipulation, it risks being treated as spam, and engines work to filter it out.
- Paid “placement in ChatGPT”. No company controls what an assistant answers.
How long does it take, and how do I know it works?
There is no official timeline. A new page can start appearing in live-search answers once it has been crawled, which may take days or weeks. What a model knows from training changes slowly, usually when its maker releases a new version. For a brand-new business, expect months, not days.
Measure with the same question set every month, repeated runs and a share you can compare. A shift of a few points between two months is noise; a steady climb over a quarter is a signal. Arcline Premium automates this loop and alerts you when an assistant stops citing you (see the features page), but nothing it does can promise a mention.