The quiet rebuild of retail: how AI changes the buying decision

01From search engine to answer engine

A search for the best ergonomic office chair used to return a results page to click through. Today many people ask ChatGPT, Perplexity or the AI overview inside the search engine instead and get an answer instead of a results list. The answer names two or three providers, the rest does not take place. The difference is fundamental: before, ten results competed for your attention, now the model decides upfront what becomes visible at all.

Nobody measures the exact share reliably. The direction is still clear: search engines place answers prominently, and where they appear, clicks on classic results drop noticeably. Product searches migrate there too: first the recommendation in the chat, then the click into a single shop.

02What this means for your visibility

AI systems cite what they can read and classify reliably. Old SEO tricks do not help with that, quite the opposite: bloated keyword texts are as useless to language models as they are to people. Visible becomes whoever delivers clear facts, appears consistent and gets mentioned credibly by others.

That shifts the work: away from chasing position one for a single keyword, towards a website that works as a reliable source. Mentions in reviews, directories and trade articles pay into that, just like pages that make their facts checkable. For small merchants this is good news rather than a threat: specialist knowledge and honest depth of detail beat pure advertising budget here.

03What machines need to be able to read

The good news: the technical foundation for this is manageable, and once set up cleanly it is quick to maintain.

• Structured data based on schema.org: product, price, availability and reviews marked up machine-readably.
• Complete product information: material, dimensions, compatibility, delivery time, concrete instead of promotional.
• Clean page structure: real headings, clear paragraphs, content reachable without script hurdles.
• An llms.txt in the root directory: a compact, machine-friendly overview of your most important content, a young but growing standard.
• Consistent facts everywhere: the same names, prices and details on your website, profiles and directories.

04Shopping agents on the horizon

The next step is taking shape: AI agents that do not just advise but search, compare and prepare purchases themselves. First payment and hand-over standards for this are emerging. As of August 2026 the everyday volume is still small, but the infrastructure agents need is the same as above: clean data, clear prices, reliable information. Whether open standards win out or a few platforms control the access is not decided yet, both cases speak for a website of your own that machines can read cleanly.

05Why none of this is rocket science

Nothing on this list is exotic. It is solid craft that already helps search engines and, above all, people: whoever wants to decide fast needs the same clear information a machine does. And if the shift runs slower than expected, you lose nothing: better product pages sell better even without AI.

If you have to prioritise: complete the product data first, then structured data, then llms.txt. That order gets you the most.