Imagine going to a restaurant and the waiter just tells you what to order.
You do not read the menu. You trust the recommendation. The waiter is not selling you anything specific. They are answering the question you actually came in to ask, which is "what should I have." The decision space collapses from twenty options to one. The friction of choosing is gone.
That is roughly what is happening right now in how people shop for consumer products online. They are no longer browsing five sites and comparing options. They are asking an AI assistant what to buy, taking the recommendation, and moving on.
For CPG brands, this is a structural change in how product discovery works. And most brands are not set up to be the one the AI recommends.
What is actually changing
For two decades, online product discovery has worked roughly the same way. The customer searches for something on Google. Google returns a list of results, ranked by some combination of relevance, authority, and ad spend. The customer clicks through several options, compares them, and eventually decides.
This model assumed the customer wanted to do the comparison work. The brand's job was to be in the consideration set, to look credible, and to win on the merits of the decision.
The model is breaking down. A meaningful and growing share of product discovery queries are now happening inside AI assistants. ChatGPT. Perplexity. Claude. Gemini. The customer types something like "what is the best gut-friendly soda" or "what is a good electrolyte drink with no artificial sweeteners," and instead of getting a list of options to evaluate, they get a recommendation. Maybe two or three options, with the reasoning attached.
The customer often takes the recommendation. They do not click through five comparison articles. They do not read fifteen reviews. They have outsourced the comparison work to the AI. The AI has done it for them, with confidence, in three seconds.
This is not science fiction. This is happening right now, at increasing scale, across a generation of shoppers who have grown comfortable trusting AI for higher-stakes decisions than what to put in their cart.
For brands that show up in the AI's recommendation, this is a phenomenal channel. The customer arrives already convinced. The conversion math is wildly favorable. For brands that do not show up in the AI's recommendation, this is invisible loss. The customer was looking for exactly what the brand sells, never saw the brand, and bought something else without ever knowing the brand existed.
How AI assistants actually decide what to recommend
There is a tendency to talk about AI search as if it is magic, or as if it operates by some opaque set of rules no one can understand. This is mostly wrong. The principles by which an AI assistant decides what to recommend are knowable, and they are not particularly mysterious.
When an AI is asked "what is the best electrolyte drink with no artificial sweeteners," it is drawing from the corpus of text it has been trained on and, increasingly, from real-time search results it can retrieve. It is looking for products that are described, in that corpus, as being in the category being asked about, with the attributes being asked about, in language that is clear and parseable.
The brands that get recommended share several traits.
The product page describes the product in plain, direct language. Not poetic brand language. Not metaphor-heavy positioning. Functional, descriptive language that an AI can extract and summarize. "An electrolyte drink mix with no artificial sweeteners, sweetened with stevia, available in five flavors." That sentence is gold. A landing page that opens with "feel the difference, hydrate like never before" is invisible to an AI looking for a recommendation.
The brand has third-party validation that an AI can find and verify. Mentions in publications the AI considers authoritative. Reviews on platforms the AI is willing to cite. The presence of the brand in lists and comparison articles published by sites the AI trusts. None of this is dishonest. It is the same kind of credibility-building that mattered in traditional SEO. The difference is that the AI is now the gatekeeper, not the search algorithm.
The product is described with structured data that AIs can parse cleanly. Schema markup on the product page. Clear, machine-readable specifications. A FAQ section that addresses the questions people actually ask, in the language they actually use, rather than in marketing-speak.
The brand has been mentioned, in some form, in conversations the AI has been trained on or has access to. Reddit threads. Quora answers. Substack newsletters. Podcast transcripts. The brand exists in the cultural conversation, not just in its own marketing channels.
Brands that miss most of these end up invisible to AI recommendations. The product can be excellent. The website can be beautiful. The brand can be doing real revenue. None of that matters if the AI cannot find a clean, structured, validated reason to recommend it.
What this means for product page architecture
The product page, for the last decade, has been designed primarily for human eyes. Hero image, brand voice, social proof, call to action. This is still important. Humans still buy products, and a beautiful product page still converts.
But the product page now has a second job. It also has to be legible to AI. And the design choices that make a page beautiful to humans are often the same choices that make it illegible to AI.
Heavy reliance on images to communicate product information. An AI cannot read a beautifully designed graphic that shows the product's benefits. An AI can read the text below the graphic. If all the benefits are in the graphic and the text is poetic brand language, the AI sees brand language. It does not see benefits.
Vague positioning language at the top of the page. "Hydration, reimagined." This is fine as a tagline. It is not fine as the primary text describing the product. An AI scanning the page wants to know what the product is. "Hydration, reimagined" is not the answer to that question.
Buried specifications. Most product pages have the actual product details, the ingredients, the size, the use case, in a section the customer has to expand or scroll to find. From an AI's perspective, anything that takes a click or a scroll to surface is downweighted. The information has to be in the primary content of the page, not in an accordion the AI may not expand.
The fix is not to make the page ugly. The fix is to make sure the primary text content of the page, the text that loads first and lives in the main body of the page, contains a clear, direct, parseable description of what the product is, who it is for, what is in it, and what makes it different. This text can live alongside the beautiful brand language. It does not have to replace it. It just has to be present.
The same principles apply to the product page work that determines whether cold paid traffic converts: communicate category, use case, and trust in the first viewport, in language a stranger can parse in five seconds.
The brands that have done this are showing up in AI recommendations. The brands that have not are not.
What this means for content beyond the product page
The product page is necessary but not sufficient. The AI's recommendation also draws on what the broader internet says about the brand, in places the AI considers credible.
This is where the work gets less direct and more durable. The brand needs to be present, in legitimate ways, in conversations and content that the AI is drawing from.
Press coverage in publications the AI considers authoritative. This is harder to engineer but it is the most powerful single lever. A mention in The New York Times, Bon Appétit, or a credible industry publication carries weight in the AI's recommendation logic far beyond its direct traffic value.
Reviews and discussion in forums and communities. Reddit threads about the category. Substack newsletters covering the brand. Podcast appearances where the founder is interviewed. These signals all feed into the AI's sense of whether the brand is real, what category it sits in, and whether it deserves to be recommended.
Comparison articles and category roundups. The kind of "best of" content that has always been a part of SEO is now a part of AI search too. If the brand is consistently included in third-party "best electrolyte drinks of 2026" articles, the AI is more likely to surface it. If the brand is never in those articles, the AI has no signal that the brand exists in that category at all.
This is not a quick fix. It is a multi-quarter discipline. Building real presence in real conversations takes time. But brands that started this work two years ago are now reaping the benefits in AI discovery. Brands that have not started are increasingly invisible.
What this does not change
There is a tendency to talk about AI search as if it is going to replace everything that came before. It is not.
Google is not going away. Direct traffic is not going away. Paid social is not going away. People will continue to find products through every existing channel for a long time, in addition to AI.
What is changing is the share of discovery that flows through AI, and the trajectory of that share. The share is small today. It is growing fast. By the time it is the dominant discovery channel, the brands that have already optimized for it will have an enormous structural advantage. The brands that are still planning to address it will be playing catch-up against competitors with a meaningful head start.
This is the same dynamic that played out with mobile-first design in 2014, and with SEO in the late 2000s. The early movers built a moat. The late movers had to spend years catching up to a baseline that was no longer a competitive advantage.
The choice in front of CPG founders right now is whether to be in the first cohort or the second.
Where to start
If you are looking at your own brand and wondering whether you show up in AI recommendations, the most useful first step is to ask. Literally.
Open ChatGPT, Claude, or Perplexity. Type the query a customer would type to find a product like yours. "What is the best non-alcoholic aperitif." "What is a good replacement for sugary sports drinks." "What is the best clean protein powder." See what comes back. See whether your brand is mentioned. See what the AI says about it if it is mentioned.
That exercise will tell you more about your current AI discoverability than any analytics dashboard will. If your brand is consistently absent from the recommendations, the work to do is structural. Product page legibility. Third-party presence. Structured data. Press and content placement.
If your brand is mentioned but with thin or inaccurate detail, the work is to improve the quality of the signals the AI is drawing from. Update the product page text. Build out the FAQ. Get the brand into more credible third-party content.
If your brand shows up in the top recommendation reliably, you are ahead of most of the category. Keep building.
The brands that win the next decade of CPG are the ones that get found when the customer asks. The customer is increasingly asking an AI. The brands that make themselves easy to recommend are the ones the AI will recommend. The rest will continue to spend on paid acquisition to compensate for discovery they used to get organically.
The waiter is going to tell the customer what to order. The brands that have done the structural work to be on the waiter's recommendation list are the ones that will be on the table. The rest will be wondering where the customer went.
