There is no ranking in AI search, not in the sense of position one and position two. There is a shortlist of names inside an answer, and either you are in it or you are not. That changes what you optimise for.
- A ranked list rewards being slightly better than the next page. An answer rewards being usable at all.
- Five things decide it: access, structure, specificity, corroboration and freshness.
- In our 2026 study of 216 Shopify stores, 41 percent failed the first one, access, which is also the cheapest to fix.
What is AI search actually doing?
It reads a question, retrieves a small number of sources it thinks can answer it, and writes a single answer from them. Ten blue links became three names and a paragraph. There is no second page to be on.
So the unit of success is not position, it is inclusion. That is a harsher pass or fail, and it is also easier to reason about, because the reasons for exclusion are concrete and checkable.
What are the five things it checks?
In the order they disqualify you.
| Check | The question it answers | Most common failure |
|---|---|---|
| Access | Can I fetch this page at all? | Crawlers blocked in robots, or content behind scripts |
| Structure | Do I know what I am looking at? | No structured data, or data that contradicts the page |
| Specificity | Is there anything here to quote? | Marketing adjectives instead of facts |
| Corroboration | Does anywhere else confirm this? | No reviews, mentions or citations off site |
| Freshness | Is this still true? | Stale prices, stock and dates |
Which one should I fix first?
Access, always, and it takes minutes rather than weeks. If assistants cannot fetch your pages, nothing else you do can matter. Check your robots file and confirm your content exists in the HTML rather than only after scripts execute, which is covered in JavaScript rendering.
After that, specificity gives the fastest visible return, because it is what makes a page quotable. Semantic density is the idea: more fact per sentence, fewer adjectives.
How do I measure whether it is working?
Not with rank tracking, which has nothing to track. Measure by asking the assistants the questions your buyers ask, recording which names come back, and repeating on a schedule so you see movement rather than a snapshot.
Track referral traffic from assistants alongside it, so you can connect being named to people actually arriving. Measuring AI referral traffic sets out how.
Common questions
Is AI search optimisation different from SEO?
It overlaps heavily. Crawlability, structure and clear writing serve both. The difference is that AI rewards being quotable and corroborated, where classic search also rewarded accumulated link authority.
Do backlinks still matter?
Yes, as corroboration. An assistant deciding whether to trust a claim looks for confirmation elsewhere, and links and mentions are how it finds that.
How long does it take to show up?
Weeks to months. Access fixes can register quickly because they change what can be fetched. Corroboration is the slow part, since it depends on other people.
Run a free scan and get a 0 to 100 score with a fix list in priority order. Run the free scan.
Your buyers are already asking AI. This is how you make your website readable to the assistants they ask.
Everything in this article is measurable on a live storefront. The Full Assessment and Roadmap reads your website the way AI crawlers receive it and hands you every fix in plain English, in the order we would make them.