A growing share of people now ask an assistant rather than searching — what's the best X, who should I hire for Y — and get an answer with a handful of sources rather than a page of results. Businesses reasonably want to be in those answers. The honest position is that some of how this works is clear and a lot of it isn't, and the advice industry has got well ahead of the evidence.
What's reasonably clear
You have to be in the index the assistant uses. Assistants don't crawl the web live in the way a search engine does; they retrieve from a search index. Several have used Bing's index, and others use different providers.
This has a practical consequence most people miss: if you only use Google Search Console, you're monitoring the wrong index. Bing Webmaster Tools is free, and for AI visibility it may matter more than Google. That's the single most actionable thing in this article.
Crawlers have to be allowed in. Assistants' crawlers respect robots.txt. If yours blocks them — deliberately or because a plugin added the rules — you're invisible to that assistant regardless of anything else. Check which bots you permit.
Content that answers a question directly tends to get retrieved. Assistants are quoting and summarising passages, so a page that answers its own question clearly, near the top, in plain language, is easier to use than one that buries the answer after six paragraphs of preamble.
Specific beats general. A page making a concrete, checkable claim is more quotable than one making a vague one.
What's genuinely uncertain
Everything that sounds like a ranking factor.
There's no published list of what makes an assistant cite one source over another. Nobody outside those companies knows the weighting, and it changes. Anyone selling you a definitive checklist is guessing with confidence.
`llms.txt` — a proposed file for telling assistants how to use your site — is a convention some people have adopted, and there's little evidence that major assistants honour it. Adding it costs nothing and shouldn't be relied on.
Structured data markup has well-established value for traditional search. Its role in AI retrieval is much less clear, though it's unlikely to hurt.
And claims about "optimising for AI" as a distinct discipline mostly repackage ordinary good practice.
What that means practically
The things worth doing are unglamorous and they overlap heavily with regular SEO.
Be in Bing's index and monitor it. Allow the crawlers. Answer the question your page is about, early and plainly. Make claims specific enough to be worth quoting. Keep information current, because an assistant quoting a stale page about your prices is worse than not being quoted.
And have the basic facts about your business stated in text on your own site — what you do, where, for whom, how to reach you. Not in an image, not only in a PDF.
How to check whether it's working
Ask the assistants. Genuinely — ask them the questions your customers would ask and see what comes back and who gets cited.
It's crude, it varies between sessions, and it's the most direct feedback available. There's no Search Console for this yet.
The thing to be sceptical about
Any service promising to get you into AI answers. Nobody can promise that, because nobody controls it and nobody outside those companies knows the mechanism. What they can do is ordinary good SEO, which is worth paying for honestly described.