The "which [product] should I buy?" query is moving from the search box to the assistant. And in an assistant there are no 40 results: there are three recommendations.
What an assistant looks at on a product page
- Concrete, comparable specs: dimensions, materials, compatibility, consumption. Adjectives do nothing; numbers do.
- Visible, current pricing, with currency and availability marked up in Product + offers.
- Reviews with volume and dates: average rating, review count and recency.
- Shipping and returns policy in plain text, not a PDF.
The pages most likely to get you cited (and almost nobody has)
- Honest first-party comparisons: "Model A vs Model B", including when not to pick yours. Models value nuance and cite it more.
- Buying guides by use case: "best [product] for [situation]". That is literally the shape of the user's prompt.
- Sizing, measurement or compatibility pages with tables.
- Per-product FAQs built from real support questions.
Third-party reviews: the external lever
Assistants lean heavily on aggregators and media. Appearing in two or three relevant industry roundups usually moves the needle more than any on-site change. Treat third-party comparison mentions as an ongoing workstream, not a one-off campaign.
Mistakes that keep you out
- Product pages with duplicated manufacturer copy: nothing of yours to cite.
- Prices visible only after adding to cart.
- Key content locked in images or video with no alt text.
- Blocking AI crawlers in robots.txt without consciously deciding to.
How to measure it
Build a set of 15-20 real buying prompts for your category, and weekly track whether your store appears, in what position and which competitors dominate. When you publish a new buying guide, watch whether your mention rate moves over the next four weeks: that tells you which format to repeat.