I did a simple experiment with my LinkedIn feed earlier this week. I scrolled ten posts deep and counted how many were AI-related.
I expected six, maybe seven. I got ten.
In rapid succession I scrolled past:
- A consulting report on how AI is already shaping how Europeans decide what to buy
- A playbook for running structured prompts on Claude for strategy analysis
- A post about the need for improving product discovery in LLMs and shopping assistants as shopping is very visual
- A guide on how to build a brand agent
- An upcoming conference about how AI is changing consumer behaviours
- An announcement about the launch of apps within an AI agents
It’s a lot. For DTC brands in particular, where acquisition costs were already rising before AI entered the conversation, the pressure to figure out how consumers will discover brands in the AI-powered environment feels especially acute.
Thankfully, there’s still time to act. While it’s easy to get drawn into the excitement and go all-in, I think we’re better off applying a structured and strategic approach to test and discover what really works.
That’s what this piece is for. To take a breath, touch grass, find the signal in all the noise, and plot a sensible course from there.
That’s what this piece is for. To take a breath, touch grass, find the signal in all the noise, and plot a sensible course from there.
However, there’s enough signal to act on.
So let’s start there.
Your customers have already moved
Before we talk strategy, let’s be clear about what’s actually happening out there. The numbers tell a clear story:
- Referral traffic from AI chatbots to US retailers increased 760% year on year in November 2025
- AI-referred visits are converting at 1.5x the rate of other sources, and more than half of those visitors are upper-funnel, net-new shoppers, i.e. people who weren’t already on their way to your site
- Over 70% of AI-referred users now land directly on product pages, up from around 50% just six months earlier
- 39% of shoppers already use AI for product discovery, and 47% use it for comparison shopping
- 84% of European consumers report using AI tools in their everyday life
That alone should be enough to make any DTC marketer pause for thought. Let’s unpack what this actually means.
I’ll use the example of a handbag with a laptop pocket I recently bought. In the pre-AI era, I would have typed “women’s handbag laptop” into Google, spent time browsing different sites to understand if it would fit my 15-inch MacBook and gradually narrowed my options down.
Each site had a real shot at winning me as a customer — the right product specifications along with product image and description. My decision was being made across those sites collectively.
Here’s what actually happened.
I started with an AI shopping assistant, sharing the context: I’m looking for a new handbag that can hold my 15-inch MacBook, to be used every day going to the office. Think feminine, structured, leather. Budget $500.
From there, the LLM asked follow-up questions, compared options, and surfaced the top five products that it found best matched my prompts. As I refined the brief, adding things like high quality, minimal design, top zipper, and explicitly avoiding briefcases, the recommendations kept improving. I even shared brands I like. Beyond the recommendations, I had direct links to subreddits and product review articles. By the time I hit a product page, my research and comparison were essentially done, and it was time to whip out the credit card.
A fundamental shift has taken place here that retailers and DTC brands need to pay close attention to.
They’ve gone from arriving to browse to arriving to buy.
The repercussions for DTC brands, and retail brands in general, are significant. Your brand either shows up in the AI-generated recommendations within these assistants recommendation or it doesn’t.
The rules of visibility have changed
Performance marketing has run on a fairly consistent logic for years: budget buys visibility. Bid higher, reach more people, win more customers. The auction model rewarded whoever had the deepest pockets in a given keyword category.
Audience targeting refined that further — define who you want to reach, build a segment, bid to win them.
AI discovery doesn’t work that way.
In simple terms, with keyword and audience-based targeting, the marketer defines the context: I want to reach women aged 30-40, mid to upper-middle income level, living in European major cities, highly connected digital-first shoppers. The platform finds people who match that description.
With conversational intent-based targeting, the consumer defines the context with far more detail than they would ever type into a search bar.
Rather than simply search for a “women’s handbag”, they’re now telling an AI they have a MacBook, actively looking for a new handbag, and a detailed brief of what they’re actually looking for.
That’s why the brand that wins isn’t the one that bid highest, it’s the one whose product is the most credible answer to what was actually asked. That’s a meaningful shift in how visibility is earned. A brand with precise, use-case-specific product data can surface ahead of a brand spending ten times more.
Let’s be honest about the limits of what we know here: the full picture of what drives organic versus paid AI results is yet to be discovered.
These are LLMs at the end of the day, so we don’t have (and may never have) full insights into how they prioritize who sees what, nor do we have enough test results currently to optimize against.
What we can say with confidence is that budget still matters — but relevance is now what opens the door.
AI notices gaps and inconsistencies in your brand story
Business of Fashion Insights and Quilt.AI tracked 28 high street fashion brands across ChatGPT, Gemini, and Claude throughout 2025. The findings were striking, and I think they extend well beyond fashion.
Brands with consistent, specific, repeated positioning showed up in AI recommendations.
Brands with diffuse, aspirational-but-vague messaging often didn’t appear at all.
The logic is fairly straightforward: AI identifies patterns across everything published about a brand online including product pages, press coverage, customer reviews, owned content, third-party commentary.
It retrieves what’s statistically clear. If a brand’s positioning is implied or inconsistent across those sources, there’s no stable pattern to find.
Take COS for example, an H&M Group-owned fashion brand built around minimalist, considered design. It came out tops for coherence ranking, meaning ChatGPT, Claude, and Gemini had a consistent understanding of what COS stands for across thousands of test prompts.
It also ranked first for prominence: when AI assistants listed brands in a relevant context, COS tended to appear near the top.
That raises a practical question worth putting to your own brand right now.
If an AI model reads everything published about you online, your product pages, press coverage, customer reviews, how accurately could it describe both what you sell, and who you are selling it to?
The most direct way to find out is to ask one.
Drop a prompt into your AI model of choice and get it to do a research exercise on your own brand, starting with a setup prompt to get useful results:
Act as a brand research analyst. Be concise and specific. Infer from messaging, product, pricing, and reviews. Focus on tensions, contradictions, and blind spots. Avoid generic brand language. Deliver structured, insight-led answers.
- Offering: In one sentence, what does [brand] actually sell? Then list core product categories, and what is commonly overlooked or misunderstood
- Customers: Who is [brand] for? Describe the core customer in terms of demographic, behaviours, needs, motivations, and who might wrongly think it’s for them
- Use cases: When do people choose [brand]? List 5 key use cases / moments, and add a few moments it likely misses
- Brand values: “What does [brand] say it stands for vs what it actually signals? Highlight concrete contradictions or tensions.”
- Consistency: “How consistent is [brand] across its website, press, and reviews? What themes align, and what conflicts?”
The gaps and inconsistencies it surfaces are exactly the ones worth addressing. Fix what you can influence: product pages and feeds, owned content, and how you respond to reviews and encourage customers to share user-generated content on social.
Press coverage is harder to control, but every channel you do own is worth improving consistency on.
Nobody has the full picture
One of the things contributing to the AI hype we’re all feeling is the fact that AI is not just moving fast, it’s also evolving in ways that are extremely difficult to predict.
Waiting for the dust to settle before writing strategy may have been a wise move in the past, but in the age of AI the dust is never going to settle.
The first-mover advantage with AI cannot be underestimated — the steps you take don’t have to be massive leaps. Weigh up your options, formulate a clear, pragmatic strategy, and move calmly forward, step by step.
Those are the challenges. They’re real, and understanding them clearly is step one.
Here’s what you can do about them.
Build your organic foundation
Step 1 is making sure AI can find you, and when it does, that it understands what you sell.
On the product data side, I could easily dedicate an entire article to AI feed optimization, but the summarised version would be: write rich, specific, use-case-led descriptions give AI systems something to match against real consumer intent.
So a product description like “versatile midi dress” becomes “a midi dress for smart-casual occasions, works well for office events and weekend dinners.”
Think about both the questions your customers are asking and your product’s use cases. Then make sure the answers and context are built into how you describe your products.
On brand narrative: brands that repeat the same values consistently, across owned content, press outreach, product pages, and customer reviews, are easier for AI to surface accurately.
Inconsistency creates noise the AI can’t resolve. And since 56% of shoppers want AI to suggest things beyond exactly what they searched for, brands with well-articulated context get more shots at relevance, not just in exact-match scenarios.
Three questions worth considering:
- If an AI read only our product pages, what use cases and occasions would it associate us with?
- Do our brand values come across the same way across owned content, press, product descriptions, and customer reviews?
- What contexts do we want to own in AI recommendations, and have we built enough consistent, repeated signal around them?
Rethink what paid is for
The old framing for paid media was simple: buy visibility, outbid competitors for attention.
In AI discovery environments, the goal is reaching net new customers whose context and intent genuinely match what you’re selling, at the moment they’re in the market for it.
That reframing changes what you measure and what you optimize toward.
Rather than asking how many impressions you bought and at what cost, the more useful questions right now are:
- What consumer groups respond best to paid discovery in AI environments?
- What does behavior look like post-click, once someone arrives on your site having come from an AI conversation?
- What role does paid discovery play across the full acquisition journey, beyond last-click attribution?
These are valuable research questions to explore. Nobody has all the answers yet, but the brands testing, proving and iterating now will be considerably further ahead as AI-driven discovery matures.
If you aren’t already experimenting, there’s never been a better time to start…
If you’re able to tune out the noise generated by all the AI hype in the market, you start to realize that though a lot of companies are testing pilots, few are using AI in a meaningful, differentiated way tied to real business outcomes.
Sure, we’ve got a shiny new box of AI tools, but we are still figuring out how to redesign workflows, business models and operations in a way that returns real value.
However, the brands at the forefront of testing advertising in AI platforms (OpenAI’s ChatGPT for example) are getting a front row seat into how consumer journeys are changing.
In summary, the two things I’d advise any DTC brand to do would be to audit your organic foundation and start testing what paid looks like in AI discovery environments.
Neither requires certainty about where things will be in two years. Both will compound in value regardless of how it develops.
There is still so much to learn and the definitive playbook on DTC discovery in the age of AI is yet to be written.
The brands best placed for whatever comes next aren’t the ones who opted for a ‘wait-and-see’ approach. They’re the ones who started moving while everyone else was still watching the ship from the shore.
The 2026 Criteo Commerce & AI Trend Report goes deeper on how AI is reshaping discovery, what shoppers want from AI-assisted experiences, and what the data says about where this is heading. It’s the research that informed a lot of what’s in this piece.
