How to ensure your fashion brand shows up in AI searches

The Knowledge Graph built a picture of your brand without your input. This will dictate whether you show up for AI searches or not.

I was doing an entity audit for a brand a few months ago for a well-run premium streetwear label, strong community and solid drops. When I searched their name in Google, the Knowledge Panel that came up described them as a sportswear manufacturer. Not fashion. Not streetwear. Sportswear. The founding year was wrong by about three years. The website listed was a redirect that no longer resolved properly.

None of their team had noticed. Why would they? The Knowledge Panel isn't part of anyone's weekly reporting. It doesn't show up in GA4. It doesn't affect the Klaviyo dashboard. Not many people even know what it is.

I'd argue it was quietly affecting everything including, increasingly, how AI shopping agents were describing and recommending their products to customers who'd never heard of them.

That's what this article is about.

What is the Knowledge Graph?

Google launched the Knowledge Graph in 2012 with a phrase that was more useful than it probably intended: things not strings.

Before the Knowledge graph existed, searching for a brand returned pages that contained brand related keywords. Afterwards, Google attempted to understand that a brand is a specific entity in the world. That entity included attributes, relationships, a category, a history and to surface related information about that entity rather than just documents containing the keyword.

This was the first step from keyword to context and we didn’t even realise at the time.

The Knowledge Graph is the database that makes that possible. It holds entities and the connections between them. Your brand, what it makes, where it operates, who founded it, what it's known for, how it relates to other entities in the same space.

The Knowledge Panel is the box that appears on the right side of search results. It is the visible output of that database. It's Google showing you what it believes it knows.

Here's the part that most ecommerce teams haven't thought about: Google doesn't build that picture from your website alone. It triangulates across multiple independent sources and weights them differently based on how much it trusts each one. Your website is one input. It's not the primary one.

How does the knowledge base get its information?

The sources Google weights most heavily, based on what I've observed across a lot of entity based work,  are roughly as follows, in descending order of trust.

Wikipedia and Wikidata sit at the top. Wikipedia because it represents independent editorial review. 

Wikidata because it's the machine-readable equivalent. Structured entity attributes that feed directly into the graph. A brand with a Wikidata entry has the closest thing to a formal identity registration that the open web offers.

Third-party editorial coverage comes next. When a credible publication writes about your brand in a specific context by placing it in a category, describing what it's known for, attributing founding details then that corroborates and strengthens the entity signal. The graph is doing something close to cross-referencing sources the way a journalist would. This is no different to how Google has always operated. Brand trust is based on what others say about you, not what you say about yourself.

Your own structured data. The Brand and Organisation schema markup on your site matters. It's weighted lower than third-party sources precisely because you're describing yourself. Google knows you have an incentive to present a particular picture. Independent sources don't. You structured data links what others say about you to your website.

GS1 registration signals legitimacy for product entities. Your Google Business Profile, social profiles, press mentions. All of these contribute, individually modestly, but collectively significantly if they're consistent with each other. Agentic trust is built through lots of signals. It’s risk mitigation from Google.

The word I keep coming back to is consistency. When multiple independent sources agree about the same entity attributes such as the brand name being spelled consistently, the founding year matches, the category description is coherent across sources then confidence in the entity goes up. When sources contradict each other, confidence goes down. The brand I described at the opening had contradictions across at least four sources. The sportswear classification was the most visible symptom, not the root cause.

Why is this so important in 2026?

For most of the past decade, the Knowledge Graph mainly affected things like how your brand appeared in rich search results, whether you got a Knowledge Panel, how Google refined searches that might refer to multiple entities. Important for SEO specialists, mostly invisible to everyone else.

That's changed. And the change is significant enough that I think it's worth being direct about what's happening.

AI shopping agents such as Perplexity, ChatGPT Shopping, and Gemini don't navigate the web the way a customer does. They don't scroll a category page, open product tabs, compare images. They query structured knowledge sources to identify entities that match the parameters of a request. They evaluate those entities for trustworthiness, and then retrieve and synthesise product information from them.

The Knowledge Graph is one of the primary sources they draw on at the entity evaluation stage. When someone asks an agent to find them a premium UK streetwear brand that does drops in a specific price range, the agent is essentially running an entity resolution query. Which entities exist in this category? What are their attributes? How confident can I be in the information about them?

A brand that exists as a coherent, well-corroborated entity in the Knowledge Graph passes that check with higher confidence. A brand with contradictory signals, thin coverage, or misclassified category attributes may not be evaluated at all.

I think this is the mechanism that explains something I've been watching in the data for several months. Brands with stronger entity signals are generating more ChatGPT referral traffic, at higher conversion rates, than brands without them. The correlation isn't perfectly clean. A lot of things affect traffic. That said it's consistent enough that I'd wager a bet on it.

The sameAs field and why most brands have it wrong

There's a specific schema property called sameAs that's the connective tissue between your self-declared entity and all the independent sources that corroborate it. You put it in your Brand schema markup, and it links your website's entity to your Wikipedia page, your Wikidata entry, your GS1 registration, your Instagram profile, your LinkedIn page.

When an agent queries the Knowledge Graph about your brand, the sameAs connections allow it to pull a coherent, multi-source picture rather than relying on your website alone. Without them, you're an island. 

The agent can find your site but can't cross-reference it against anything. It cannot trust you.

I've seen brands where sameAs was pointing to an Instagram handle that was changed two rebrands ago. 

I've seen it linking to a Wikipedia disambiguation page rather than a brand article.

I've seen it absent entirely. The Brand schema is there, the sameAs property just never got populated, probably because whoever set up the schema didn't fully understand what it was for.

It's one of those things that lives in the gap between technical SEO, brand management, and whoever manages the Shopify theme. Nobody explicitly owns it, so nothing gets done. I'd guess maybe a third of the fashion brands I've worked with have accurate sameAs implementation. That's being generous.

Why category positioning is so important

The entity confidence problem has two dimensions. One is whether the agent knows you exist and trusts the basic facts about you such as name, website, founding date, sector. The other is whether it understands what category you belong to and what you're known for within it.

These are different problems with different solutions.

The first is addressed by entity registration work. Wikidata entries, sameAs links, schema consistency, Knowledge Panel verification. The second is addressed by what I'd call category content: long-form authoritative writing about the areas your brand leads in, structured to teach the graph what you're known for.

For a heavyweight streetwear brand operating a drop model, the graph knowing you exist is necessary but not sufficient. It also needs to understand that you belong in the category of premium UK streetwear, that you're known for product weight and construction quality, that your commercial model is drop-based rather than always-available. Those are the attribute signals that determine whether you appear when an agent is asked a category query rather than a brand-specific one.

The way the graph builds that understanding is through consistent category language across editorial sources such as press coverage that describes you in specific terms, your own long-form content that stakes clear category authority, review content from customers that uses the product language the brand uses. It's less structured than the entity registration work and harder to measure, but I'd argue it's ultimately more commercially significant. Brand-specific queries will find you anyway. Category queries are where the new customer acquisition happens.

The warning for brands who do not do this

Here's what I think most brands don't fully appreciate. The Knowledge Graph is not a static snapshot. It's continuously updated as new sources are crawled and new signals are processed. This is the important part. It's also influenced by the training data that underlies the AI models making recommendations.

Large language models were trained on data that included Knowledge Graph-corroborated information. 

Entities that were well-represented in the graph at training time are better understood (trusted) by the model. Entities that weren't are handled with more uncertainty. The model may know the name but have thin or inaccurate attribute data for it.

This creates a compounding dynamic. A brand with strong entity signals was better represented in training data. Better representation in training data means higher baseline confidence in AI responses about it. 

Higher confidence means it surfaces more readily in agent recommendations. More recommendations mean more editorial mentions and citations. Those citations feed back into the Knowledge Graph and improve entity signals further.

The inverse is also true. A brand with weak entity signals gets recommended less, generates fewer citations, sees its graph signals stay thin, and falls further behind as competitors compound their advantage.

I'm still working out exactly how long this compounding cycle takes to manifest in measurable traffic. I'd estimate 12 to 18 months is the relevant horizon based on what I'm seeing. Which means the brands doing this work now are building advantages that will be visible by mid-2027. The brands starting in 2027 will be catching up to a moving target.

This work needed to have been done 12 months ago. It needs picking up now if it hasn’t been started.

What do you need to do in 2026?

I want to be specific here because the entity trust conversation can slide into abstraction pretty quickly and become useless for the person who has to actually run the work.

The first thing is a Wikidata entry. If your brand doesn't have one, it should. This is not Wikipedia. You don't need editorial notability. Wikidata accepts any entity that can be defined with verifiable attributes. Brand name, founding date, country, official website, industry, notable products. It takes about an hour to create and it's probably the highest-leverage single action in the entity trust stack for a brand starting from scratch.

The second is sameAs accuracy. Pull your Brand schema using Google's Rich Results Test or validator.schema.org. Check every URL in the sameAs array. Is the Instagram link pointing to your current handle? Is the Wikipedia link going to a brand-specific article or an unclear page? Is the Wikidata link there at all? Fix everything that's wrong.

The third is Knowledge Panel verification. Search your brand name. If a panel appears and displays a 'Suggest an edit' or 'Claim this Knowledge Panel' option, do it. The claim process verifies you as the authoritative source and gives you the ability to correct inaccurate information. If no panel appears, the entity signals aren't strong enough yet. The Wikidata and sameAs work will help over time.

The fourth is category content. This is the longest horizon work and the hardest to attribute directly. Write long-form content about the areas your brand leads in. Not product descriptions, not campaign copy, but authoritative writing about the category you occupy. Heavyweight cotton construction. The drop model as a commercial strategy. Premium streetwear positioning in the UK market. This content teaches the graph what you're known for and provides the editorial corroboration that category query ranking depends on.

None of this is technically complex. It's mostly consistent, slightly tedious work that keeps getting deprioritised because it doesn't show up in the metrics anyone looks at every week. Which is exactly why most brands haven't done it. This is exactly why it's still worth doing.

The Knowledge Graph has been building a picture of your brand since you launched. It pulled from whatever editorial sources existed, reconciled contradictions as best it could, made category judgements based on available signals. You probably didn't have much input into that process.

The difference now is that AI shopping agents are querying that picture when they decide whether and how to recommend your products to customers who've never heard of you. The picture being accurate, coherent, and authoritative about the right category is no longer just an SEO consideration. It's a customer acquisition consideration.

It's not the most glamorous part of ecommerce operations. But for brands that rely on discovery, new customers finding them through recommendation rather than paid search, it's becoming one of the more consequential ones.