Being recommended by AI now depends on product data, product pages, PR and measurement working together. Here's what we brought back from BrightonSEO, and what it means for consumer brands.
Strong SEO no longer guarantees a place in AI answers. That was the clearest message from the 13 talks our CCO, Lucy Davis, attended at BrightonSEO this October as we continue to build out our AEO consultancy.
AI search draws on brand, product data, PR and measurement at once. No single team owns all of that, so no single team can win it alone.
Brand demand and third-party evidence count for far more than on-site optimisation. Across one study of 250 retail sites, brand demand tracked AI visibility closely while schema, headings and alt text barely registered. When a brand appeared in an answer to a prompt that didn't name it, the source was almost always someone else's website.
For a consumer brand, that means picking one clear message and repeating it everywhere, then investing in being talked about. A useful test: if the website disappeared tomorrow, would AI still have enough to recommend the brand?
Our first takeaway was about people. The most useful people in AI search understand more than one discipline, and the brands doing well have teams that work much more closely together.
The prompts people type into ChatGPT are intent data, and they're as useful to product, creative and PR teams as they are to SEO. PR works best when it starts from the business goal and the pages that need to move, with the campaign idea coming last. None of that happens if specialists only talk to each other.
It also doesn't happen if everyone's week goes on pulling reports together. AI is very good at the repetitive parts of this work, like prompt monitoring and monthly reporting. Brands should hand those over, then spend the time saved talking to their teams and planning how to hit their goals. Start with one repetitive task, build the smallest fix that works, and keep a person checking the output.
In Europe, ChatGPT Shopping still pulls products from Google Shopping - amongst other indexes - so a brand's feed heavily impacts how it shows up in AI shopping results. Today, most product feeds fill in the required fields and leave the optional ones empty. As one presenter put it: "Required gets you listed. Optional gets you recommended."
The recommended fixes are practical - and if we’re being honest, still experimental. Fill in optional, conversational attributes like product highlights and Q&As, put the most important terms in the first 40 words of each description, and list every size and variant as its own item. Performance marketing teams already own this work. It now counts as AEO too.
Shoppers increasingly land straight on a product page after an AI assistant has done the research, so that page has to answer their questions on its own. Most AI crawlers won’t bother to read content that loads with JavaScript, which means size guides, reviews and product details loaded that way are invisible to them.
The fix is to render that content on the server. From there, add sections that answer real buying questions ("perfect for", "when to use"), and make filter and collection pages indexable so AI can find a page that matches the prompt. Storing product attributes once, in Shopify metafields or similar, keeps the product page and the feed consistent, which matters because Google checks them against each other. This is where engineering and performance marketing need to plan together.
Our second takeaway came out of the digital PR sessions. Exclusive information is irresistible to journalists. Fact-based, data-driven journalism, published by authoritative titles and containing information nobody else has, is irresistible to LLMs.
Editorial "best of" roundups are among the sources AI leans on most for product questions, so earning a place in them is now a direct route to being recommended. And a brand can't be called the best by AI if it's the only one saying so.
Consumer brands sit on plenty of data that feels ordinary to them and is new to everyone else, such as return reasons and review text. The opportunity is to find the story in it and back it with a methodology a journalist can defend. The publications to pitch are the ones AI already cites in the brand's category. Run the same method again across new categories and it keeps producing stories.
Mention rate, the headline number on most AI visibility dashboards, is close to a vanity metric. Being cited as a source and being recommended to a shopper are different outcomes, and only one of them sells products.
Standard analytics also miss most of AI's effect. People research in an assistant, then come back later through a branded search or by typing in the URL. Branded search, direct traffic and asking customers where they heard about the brand all belong in AI search reporting. Report what the work influenced in commercial terms, rather than claiming it caused every sale.
There are plenty of AI visibility tools, and they're useful for tracking. They can't tell a brand what to fix first. A famous brand with thin product data and an unknown brand with perfect product pages need opposite things. The better question is what's holding the brand back most right now, and answering it takes someone who knows its margins and customers.
Timing matters too. OpenAI's own shopping system and Google's checkout inside AI Mode are US-only for now. Lucy expects Google's checkout to reach the UK and Europe before long, and brands with complete feeds and readable product pages will be ready when it does.
Everyone, which in practice often means no one. Work that belongs to every team tends to fall between them.
We connect engineering, design, performance marketing and AI in one team, so the product feed and machine-readable product pages are already part of our day-to-day work. Brand, PR and measurement shape how far that work goes, so we plan it with the people who own them, and use AI to clear the repetitive work so there's time for those conversations.
Ranking on Google still matters. It just won't get a brand into the answer on its own.
Thank you to all the wonderful speakers at BrightonSEO whose presentations helped shape this article: Adam Preston (Summit), Jean Sarunporn (MG OMD), Rosa Mitchell, Tom Stanton (Anything Is Possible), Harriet Stanger, Christian Goodrich (SOZO), Malte Landwehr (Peec AI), Stijn Bergmans (Follo), Greta Koivikko (NOVOS), Amy Gibson (Digitaloft), Madeleine Lambert (Originality.ai), Paris Childress (GEOforge), Samanyou Garg (Writesonic), Tom Wells (Vaer) and Cosmin Negrescu (SEOmonitor).