AEO for ecommerce: how products get recommended in AI answers

Two kinds of product answer

When someone asks an assistant about products, they get one of two things, and they are won differently.

Product cards — a carousel of items with prices and merchants. These are driven by structured product data: feeds, product markup and the shopping systems that aggregate them. Here, accuracy of price, stock and identifiers decides whether you are eligible at all.

Written recommendations — “for a small apartment, the best options are…” with sources cited underneath. These are driven by the same retrieval as any other answer: the roundups, comparison sites, reviews and forums the engine pulls for that question. Here, being mentioned on the right third-party pages decides whether you are named.

Most stores only work on the first. The second is often where a considered purchase is actually decided.

Platform by platform, for an Australian store (October 2026)

Platform	How products are chosen, in the vendor’s words	Status for Australian merchants	What to do now
ChatGPT	Results “selected independently by ChatGPT”; merchants ranked on “availability, price, quality, and whether they are the maker or primary seller”; structured metadata from first- and third-party providers	Shopping “currently live in the U.S.”; feed onboarding for approved partners, standard upload targets the US; Instant Checkout launched US-only in September 2025 and was reportedly moved into retailer apps in March 2026	Allow OAI-SearchBot; make product pages complete and crawlable; apply via OpenAI’s merchant page
Google AI Mode and AI Overviews	Shopping Graph of “more than 50 billion product listings”, about 2 billion refreshed every hour; Google’s AI-features guidance includes keeping Merchant Center up to date	AI Mode live in Australia since 8 October 2025; Google said in May 2026 its AI checkout would roll out to Australia “in the coming months”, with major Australian retailers reported as launch partners	A complete, accurate Merchant Center feed with free listings on; Product markup that matches it
Perplexity	Product cards it describes as “not sponsored”; a free merchant programme for sharing product data	Buy with Pro and the PayPal-powered checkout launched for US users	Allow PerplexityBot; the merchant programme matters mainly if you ship to the US

The pattern is clear: the transactional layer is being built in the US first, and Google is the platform with stated plans for Australia. The discovery layer — being read, retrieved and named — is open to Australian stores today on every platform.

Does AI shopping traffic actually sell?

The evidence is genuinely mixed, and it is worth knowing both sides before spending.

For. Adobe, measuring visits to US retail sites, found AI-referred traffic up 393% year on year in the first quarter of 2026 and converting 42% better than other traffic in March 2026 — a reversal from converting 38% worse a year earlier. Adobe also found product pages were the least machine-readable part of retail sites, scoring 66% against 75% for homepages, which is a fixable problem.

Against. A University of Hamburg and Frankfurt School study of 973 online shops found ChatGPT referrals were under 0.2% of sessions and converted worse than organic search. Contentsquare’s benchmark across 6,500 sites found AI-referred traffic converting at 1.3% against 1.9% for email. Even AirOps, whose data is otherwise favourable, puts the conversion edge at only 1.3× in low-consideration ecommerce.

In Australia. Australians spent a record $82.6 billion online in 2025, up 14%, according to Australia Post, which forecasts that agentic AI could influence up to 30% of ecommerce transactions by 2030 — a forecast, not a measurement. A 2026 survey of 1,000 Australian online shoppers found 20% already used AI tools when shopping, but only 4% wanted AI help at the checkout. Australians are using assistants to research and compare, not yet to buy.

What AEO looks like for an online store, in order

1. Fix the product pages. Price, stock, shipping, returns, materials, dimensions and identifiers, stated in plain text on the page and in matching Product markup. If an engine cannot read your price and availability, it cannot rank you on them.
2. Get the Merchant Center feed clean. Free listings on, no disapprovals, prices and stock that match the site. It is the input to Google’s shopping layer, and the one with Australian checkout on the way.
3. Be the primary seller where you can. OpenAI names “whether they are the maker or primary seller” as a ranking factor. If you resell, make sure your listings are not weaker copies of the brand’s own.
4. Collect genuine reviews. OpenAI says its review summaries are drawn from public websites. The ACCC’s rules apply: no fake reviews, incentives regardless of rating and disclosed, no suppressing negatives.
5. Get onto the roundups. For written recommendations, find the “best X in Australia” lists, comparison sites and review publishers the engines cite for your category — the citation map does this — and earn a place on them.
6. Allow the crawlers and measure. Allow OAI-SearchBot, PerplexityBot, Googlebot and Bingbot, then run product-shaped prompts — “best standing desk for a small apartment under $600” — on each engine, monthly. How to get cited by ChatGPT covers the crawler details.

If your margins are thin, read is AEO worth it? before spending anything; and the free check will show in five minutes whether the engines name your store today.