Can We Help Fashion Brands Get AIO Right?

Yes, we can help a fashion brand get AIO right, in specific ways: we measure how AI answers name, rank, describe and source a brand against its rivals, and turn what we find into fixes the brand can own. Nobody can promise more, because no outside party controls what ChatGPT, Gemini, Claude, Perplexity or Google's AI Overviews say. AIO, short for AI optimisation and also called GEO (generative engine optimisation) or AEO (answer engine optimisation), is the work of getting a brand found, described accurately and recommended in AI answers.

We speak from experience: McLeuker AI is a young brand that has to be found in the same AI answers, so everything below is something we have had to make work on our own site first.

Key takeaways

• AIO is mostly search and brand discipline. Google calls optimising for its AI features "still SEO" and says llms.txt files, special markup and rewriting pages for AI are not needed.
• AI answers are assembled from what is published. A brand can shape that evidence: product pages and feeds, retailer listings, press, reviews and reference sites. It cannot control the answer.
• For fashion, product data decides shopping answers. Size, fit, material, care, price and availability belong in structured data and in plain text, identical on every channel.
• Sustainability claims are now a legal question too. In the EU, the rules against generic environmental claims a brand cannot prove apply from 27 September 2026.
• Measure share of answers, not one screenshot. Answers vary from run to run, so the unit is a panel of questions asked many times, on several engines, against named rivals.

What is AIO, and how is it different from SEO?

AIO (AI optimisation) is the practice of making a brand easy for AI systems to find, understand, describe correctly and recommend. The engines are the assistants and AI search features shoppers now ask: ChatGPT, Google's AI Overviews and AI Mode, Perplexity, Claude, Gemini and Microsoft Copilot.

AIO is not a separate discipline from SEO, and Google says so. Its May 2026 guide, Optimizing your website for generative AI features on Google Search, calls optimising for generative AI search "still SEO" and lists llms.txt files, special markup, chunking content and rewriting pages for AI among the things that are not needed. Google's page on AI features and your website adds that there are no additional requirements to appear in AI Overviews or AI Mode beyond being indexed and eligible to show a snippet.

What changes is what a brand wins and how it loses.

How do AI engines decide what to say about a fashion brand?

AI engines decide what to say about a fashion brand from two inputs: what they retrieve when the question is asked, and what the model learned in training. Retrieval dominates anything current, such as prices, stock and recent press; model knowledge shapes the default picture of what the brand is known for.

Retrieval works like several searches at once. Google says AI Overviews and AI Mode may use "query fan-out", issuing multiple related searches across subtopics and data sources to build one answer (Google Search Central), so a brand that surfaces for fit or price alone can become a candidate.

Shopping answers lean on product data. Google said in May 2025 that its Shopping Graph, which AI Mode shopping draws on, held more than 50 billion product listings with details such as reviews, prices, colour options and availability (Google). OpenAI says ChatGPT's product results are chosen independently of ads, and that the merchants listed for a product come from third-party metadata or from merchants directly (OpenAI Help Center).

Third-party sources carry real weight. In an Ahrefs study of 75,000 brands, branded web mentions had the strongest correlation of any factor measured with how often a brand appeared in Google's AI Overviews, at 0.664, and Ahrefs is clear that correlation is not proof of cause (Ahrefs).

Answers also vary. In research by SparkToro and Gumshoe published in January 2026, 2,961 prompts were run repeatedly through ChatGPT, Claude and Google's AI; the same list of brands came back in fewer than one run in a hundred, while how often a brand appeared was far steadier than its position (Search Engine Journal).

And in fashion, being known is not the same as being recommended. BoF Insights analysed 120 luxury, high street, contemporary and sportswear brands across ChatGPT, Claude and Gemini between January and July 2026. A brand can be well known to a model yet rarely recommended, recommended for the wrong products, audiences or occasions, or named consistently but placed behind competitors (The Business of Fashion, October 2026).

What do we know about AI referrals and clicks?

AI answers send fewer clicks per search, while the visits they send to retailers are growing quickly from a small base. Every figure below carries its source and date, because this area moves every quarter.

• Fewer clicks when an AI summary appears. Pew Research Center tracked the browsing of 900 US adults in March 2025. Users who saw a Google AI summary clicked a traditional result in 8% of visits, against 15% without one, and clicked a link inside the summary in about 1% of visits (Pew Research Center, July 2025).
• Being cited still counts. Across 3,119 informational queries from 42 organisations, Seer Interactive found that brands cited in an AI Overview earned 35% more organic clicks than brands that were not, while noting it cannot prove the citation caused the difference (Seer Interactive, 2025).
• AI referrals to retail are rising from a modest base. Adobe reported traffic from generative AI tools to US retail sites up 693.4% year on year over the 2025 holiday season, and said the base of users remains modest (Adobe, January 2026). For the first quarter of 2026 it measured AI traffic to US retail sites up 393% year on year, and found that major portions of retail websites are not fully readable by machines (Adobe, April 2026).

None of these numbers tells a brand how often it is named in answers that produce no click, which is why AIO measurement starts from the answers, not from analytics.

What do fashion brands get wrong about AIO?

Most AIO mistakes in fashion come from treating AI answers as a new channel to game rather than a new reader of the same evidence.

• Chasing single prompts. A screenshot of one ChatGPT answer is an anecdote. Rewriting pages to win one prompt is optimising noise.
• Writing for robots. In the GEO study, keyword stuffing scored below leaving the text alone, and Google lists chunking and rewriting for AI among the things it does not need.
• Leaving crawler access to someone else's default. A CDN setting, a platform update or a security app can block AI crawlers without anyone at the brand deciding it; Cloudflare began blocking known AI crawlers by default for new domains in July 2025 (Cloudflare). The opposite mistake is blocking every AI user agent, including the search bots that decide whether a brand can be cited.
• Treating llms.txt as a lever. llms.txt is a proposal published in September 2024 (llmstxt.org), not an adopted standard. It is cheap to publish; it is not a strategy.
• Letting brand facts drift. When the About page, a wholesale listing and a reference article disagree on the founding year, the price tier or where the product is made, an answer engine has to pick one.
• Making claims that cannot be proven. "Eco", "conscious" and "responsible" ranges are exactly the copy an assistant repeats when asked whether a brand is sustainable.
• Manufacturing mentions. Fake consumer reviews are a banned practice under EU consumer law (Directive (EU) 2019/2161) and under a US Federal Trade Commission rule in force since October 2024 (Hogan Lovells).

What is a practical AIO playbook for a fashion brand?

A practical AIO playbook has six parts: product data, brand facts, substantiated claims, a crawler decision, earned sources and measurement. None of it is new work; what is new is that one AI answer can expose the gaps in all six at once.

1. Make product data complete, structured and identical everywhere

• Mark up product pages with schema.org Product and Offer. Include price, availability, brand, colour, material, size, GTIN, shipping and returns, and group variants so each colour and size is its own item (Google Search Central). Markup must match what the page shows; it is a route to the facts, not a shortcut around them.
• Keep the merchant feed complete. Google Merchant Center asks apparel offers for gender, age group and colour, and shoes for size as well (Google Merchant Center Help). Google's AI guide names Merchant Center feeds as a way for products to be visible in AI responses, and OpenAI accepts product feeds through its merchant programme.
• Write size, fit and care as text. A size chart saved as an image or composition shown only in a photo of the care label is hard for a text-reading system to use. "Runs small, size up for a relaxed fit" is a sentence an answer can repeat.
• Make the page agree with the label. Fibre names and percentages should match the sewn-in label and every retailer listing; in the EU they are set by Regulation (EU) No 1007/2011.
• Publish real market versions. Shoppers ask assistants in their own language, so each market needs pages in its language, linked with hreflang (Google Search Central), with local prices and delivery terms.

Product data drifts where a team re-keys one style into several systems. Our Product Data Agent takes a style from the supplier sheet to a checked listing on a Shopify store, with fibre names checked against EU rules.

2. Keep brand facts identical across every source

Brand facts are the identity an answer engine has to resolve: name, founding year, founders, headquarters, ownership, categories, price tier, where the product is made and main stockists.

• Write one canonical version on the About page, and state it in Organization markup with sameAs links to the brand's official profiles.
• Send the same facts to partners. A retailer's product page may be the page an answer cites for your product.
• Check the reference sources. If the brand has a Wikipedia article or a Wikidata item, make sure it is accurate. Wikidata has its own notability policy, and Wikipedia's conflict of interest guideline means a brand should propose corrections on the talk page rather than edit its own article.

Our own Organization markup carries our official profiles and our French registry identifiers, and deliberately leaves out a VAT identifier, because our legal notice states none and a computed one would be a guess. A fact that does not resolve where it claims to is worse than an absent one.

3. Only publish claims you can substantiate

Sustainability copy is where AIO and regulation meet. An AI answer that repeats an unsupported claim spreads the exposure; a specific claim, such as a named certification with its scope, is the kind of fact an answer can quote.

• The EU rules changed on 27 September 2026. From that date, EU member states must apply Directive (EU) 2024/825 on empowering consumers for the green transition (EUR-Lex). It bans generic environmental claims such as "eco-friendly" or "green" where the trader cannot show recognised excellent environmental performance, sustainability labels not based on a certification scheme or set by public authorities, and claims of a neutral, reduced or positive environmental impact based on greenhouse gas offsetting.
• The Green Claims Directive is stalled, not in force. The separate proposal, which would have required claims to be verified before publication, has been stalled since June 2025, when the European Commission announced it intended to withdraw it (CMS).
• The UK regulator has already acted on fashion. In March 2024 the Competition and Markets Authority secured formal undertakings from ASOS, Boohoo and George at Asda over the environmental claims made for their fashion ranges (CMA).

For the wider EU compliance load, read our ESPR and CSRD guide.

4. Make a written decision about AI crawlers

AI crawlers do three different jobs: indexing pages for an answer engine's search, fetching a page when a user asks, and collecting content that may train future models. Blocking a training crawler and blocking a search crawler have very different consequences.

Sources: OpenAI, Google, Anthropic and Perplexity.

Write the decision down, date it, and check it against what the site actually serves. Our robots.txt allows search bots, user-triggered fetchers and training crawlers as three documented classes, a deliberate choice for a young brand that wants to be both cited and known. Rewriting it in September 2026, we found that giving each named crawler its own group had detached every one of them from our Disallow rules, because a crawler that finds a group naming it ignores the catch-all group (RFC 9309). The file had read correctly for months; an automated check now holds its shape on every build.

5. Earn the sources answers cite

Answers cite the places that write about a brand, so part of AIO is the old work of earning coverage: press, editorial roundups, stockists, reviews and community discussion.

• Use citations as a target list. The sites an AI answer cites when it recommends a rival, and never when it recommends you, are the clearest PR brief a brand can get.
• Make your own pages worth quoting. In the GEO study by Aggarwal and colleagues, adding quotations, statistics and cited sources raised a source's visibility in generative engine answers by up to 40% in the authors' tests (KDD 2024). That is a benchmark, not a guarantee; Google's guide points the same way, toward original, non-commodity content.
• Treat reviews as product data. Genuine reviews that discuss fit, fabric and durability answer the questions shoppers ask assistants.

6. Measure it, and keep measuring

1. Write a question panel from what shoppers actually ask: the best brand for a product at a price, comparisons with named rivals, whether the brand runs small, whether it is sustainable, where to buy it.
2. Run every question repeatedly on each engine, and record whether the brand is named, its position, how it is described, which rivals appear and which sources are cited.
3. Score share of answers, not rank, and check the description as well as the mention: a brand named for the wrong product or an old claim has an accuracy problem.
4. Use the platform reports that exist. Search Console counts AI Overviews and AI Mode within overall Search traffic without separating them (Google Search Central), and Bing Webmaster Tools added an AI Performance report in February 2026 showing how often Copilot and Bing's AI summaries cite a site's pages (Bing Webmaster Blog).
5. Repeat on a schedule, monthly or quarterly, and after every launch or site change, so the trend means something.

Where does McLeuker AI help, and where does it stop?

McLeuker AI helps with the measurement and with turning findings into work a team can own. It does not sell placements, write reviews or promise an answer.

• Brand Omniscan measures AI visibility today. The Brand Mapping scan in Brand Omniscan puts 11 questions a shopper would ask to the AI answer engines it is configured for: answers that search the live web, answers from a model's own knowledge, and Google's AI Mode. It reads which brands each answer names, ranks and praises, measures visibility, share of voice, position and sentiment per engine against up to nine rivals, and lists the sites answers cite for a rival and never for the brand.
• It audits the website the way search and AI engines read it. Beside its rivals' sites, the brand's own site is checked for which crawlers robots.txt lets in, product data, market versions and social profiles, with a fix list that says what each failure costs.
• It reads the claims. The Sustainability Scan reads the brand's sustainability page and up to three rivals' for certifications, dated targets and generic green claims, checks claimed schemes against their own directories, and lists the claims the EU's green-claims rules reach.
• It turns findings into a plan. The Execution Roadmap gathers what earlier scans hand over, including website fixes and the questions and sources rivals win, into one action register with an owner, a priority and a window in the first 90 days.

Each scan runs against the live web from a brand name or URL, says how many searches it takes before it starts, and is kept against the brand so it can be reopened and run again. The Brand Omniscan guide walks through every scan, and brand intelligence shows how it fits a brand team's wider competitive work.

Where it stops is just as clear. A scan reads the public web, so it is only as complete as what has been published, and the compliance and sustainability readings are not legal advice. Measurement alone changes nothing a model says: the brand still has to fix the data, the facts, the claims and the coverage. What we bring to that is the discipline we run on our own site, down to FAQ markup generated from the text the reader sees.

Frequently asked questions

What does AIO mean for a fashion brand?

AIO, or AI optimisation, means making a fashion brand easy for AI answer engines such as ChatGPT, Google's AI Overviews, Perplexity, Claude and Gemini to find, describe accurately and recommend. The work is mostly complete product data, consistent brand facts, substantiated claims, a deliberate crawler policy, earned coverage and regular measurement.

Is AIO different from SEO?

AIO overlaps heavily with SEO: Google describes optimising for its generative AI features as "still SEO". The difference is the outcome. A brand wins a mention and a description inside an answer, so it is measured by the share of answers that name it, not only by rankings and clicks.

Can anyone guarantee that a fashion brand appears in ChatGPT or Google's AI Overviews?

No. No agency, tool or platform controls what an AI model says, and answers vary from one run to the next. A brand controls the evidence those systems read: its product data, facts, claims, crawler access and coverage.

Should a fashion brand block AI crawlers?

A fashion brand should decide crawler by crawler. Blocking search bots such as Googlebot, OAI-SearchBot, Claude-SearchBot or PerplexityBot removes the brand from those engines' answers. Training crawlers such as GPTBot and Google-Extended can be blocked separately without affecting inclusion in ChatGPT search or Google Search.

Does llms.txt help a brand appear in AI answers?

No major AI answer engine has confirmed that it uses llms.txt to decide what to cite, and Google says it is not needed for its AI features. llms.txt is a September 2024 proposal, cheap to publish and read by some developer tools, but no substitute for product data, consistent facts and earned coverage.

How do you measure a fashion brand's visibility in AI answers?

Run a panel of questions shoppers actually ask, repeatedly, on each AI engine, against named rivals. Record whether the brand is named, its position, how it is described and which sources are cited, and track the share of answers that name it over time.

How does McLeuker Brand Omniscan measure AI visibility?

The Brand Mapping scan in McLeuker Brand Omniscan asks AI answer engines 11 questions a shopper would ask and reads which brands they name, rank and praise. It measures visibility, share of voice, position and sentiment per engine against up to nine rivals, lists the sources answers cite for rivals and never for the brand, and audits the brand's website with a fix list.