An e-commerce client running a small home-goods DTC brand called me in a bit of a panic. Her top-performing blog post — a genuinely useful "our humidifier vs. the leading competitor" comparison — had been holding a page-one spot for over a year and quietly feeding her product page real traffic. Then, over about six weeks, the clicks on that post cratered. Impressions in Search Console barely moved. She hadn't been deranked. Nothing on the page had changed. What had changed was that when someone typed the same comparison into ChatGPT instead of Google, they got a full answer — spec table, a recommendation, a couple of caveats — without ever landing on her article or her product page. The search hadn't disappeared. The click had.
That's the conversation this piece grew out of, because it's not a one-off. It's the same pattern behind every "why did my comparison content stop converting" question I've fielded from product-based brands this year, and it deserves a straight answer instead of a shrug.
You can't buy your way into an AI engine's comparison answer, and you can't force a click the way a blue link used to guarantee one. What you can do is make your own product page the single clearest, most specific, most verifiable source on the comparison — so when an engine does need to cite something, or a shopper does click through to confirm the answer, your page is there holding up.
What's actually happening when AI answers instead of linking
When a shopper asks ChatGPT, Perplexity, or Google's AI Overviews to compare two products, the engine isn't running a search and handing back ten blue links — it's synthesizing an answer from whatever pages it has crawled and judged trustworthy enough to cite, then presenting a direct recommendation inside the conversation. Google has been explicit that AI Overviews are designed to reduce the need for follow-up searches, and its own AI features documentation describes surfacing a synthesized answer with supporting links, not requiring a click to complete the task. Similarly, OpenAI's and Anthropic's crawlers — GPTBot and ClaudeBot — are built to fetch and index page content for exactly this kind of synthesis, not to drive traffic back to the source in the way a search results page does.
For an informational query that used to be answer enough at the ten-blue-link stage, this barely registers as a change. For a comparison query sitting right above a purchase decision, it's existential — because the entire commercial value of ranking for "X vs Y" content was always the click that followed the answer, and that click is what's disappearing.
Where the engine actually pulls its answer from
This is the part most DTC brands get wrong first: they assume the engine is reading their own comparison page, when in most cases it's stitching the answer together from third-party sources — Reddit threads, independent review sites, Wirecutter-style buying guides, and other brands' own comparison pages — rather than either product's official page. If your product's official spec sheet lives behind a PDF, a slider-based configurator, or a page with no structured data, an engine has an easier time pulling from a review site that already laid the facts out in plain, quotable text. The brands losing this game aren't losing to their direct competitor. They're losing to whichever third party made the facts easiest to lift.
This is also why reviews drive AI recommendations more heavily than most brands expect: third-party review volume and sentiment are exactly the kind of independent, cross-referenceable signal an engine uses to decide which product to actually recommend once it has the specs. A brand that ignores its own review profile while polishing product copy is optimizing the input the engine trusts least.
What actually moves the needle: build the source page yourself
The single highest-leverage move for a DTC brand is publishing an honest, specific comparison page in your own voice — not a landing page dressed up as content, but a real teardown that names the competitor, states real specs, and says plainly where the competitor wins. See our guide to writing a comparison page AI actually cites for the exact structure. The reason this works is not persuasion — it's that a single well-structured page answering "X vs Y" directly gives an engine one clean source instead of forcing it to assemble the answer from five scattered pages, and engines have a documented preference for reducing that assembly work. A page that hedges, oversells, or omits the competitor's real advantages reads as marketing and gets treated as marketing — discounted or ignored.
- Name the competitor explicitly. Vague "leading competitor" language gives an engine nothing to match against a real query. Use the actual product name.
- State exact specs, not ranges. "10.5 hours of battery life" beats "long battery life" every time a machine has to decide what to quote.
- Concede a real weakness. One honest "if you need X, the competitor is the better pick" sentence does more for citation trust than ten paragraphs of self-praise.
- Mark it up as a product with structured data. Use Product schema with price, availability, and aggregate rating so the facts are machine-readable, not just visually laid out. See our guide on optimizing product pages for AI recommendations.
The teardown: what a citable e-commerce comparison page has that most don't
Pull up ten DTC comparison pages and most share the same failure pattern: a spec table with vague adjectives instead of numbers, no mention of the actual competitor by name (legal caution dressed up as strategy), and a "buy now" CTA sitting where an honest caveat should be. Contrast that with a page structured for citation: a spec table with numbers and units, the competitor named in the H1 or H2, one paragraph that states where the competitor genuinely wins, and Product schema plus Review schema in the page's JSON-LD so an engine doesn't have to infer the price or rating from a rendered layout.
| Element | Typical DTC comparison page | Citable comparison page |
|---|---|---|
| Competitor naming | "a leading competitor" | Named directly, by product |
| Specs | Adjectives ("great battery life") | Exact numbers with units |
| Weaknesses | Omitted or buried | Stated plainly, once |
| Structured data | None or thin | Product + Review schema |
| CTA placement | Interrupts the comparison | Sits after the honest answer |
None of that requires new inventory, new photography, or new pricing — it's a rewrite of a page that, in most catalogs, already exists in weaker form.
Fix the review layer, not just the product page
Because third-party review volume and sentiment carry real weight in how an engine resolves a comparison, the highest-leverage work outside your own site is making sure your review profile is current, substantial, and honestly represented — on your own site with AggregateRating schema, and on the third-party platforms an engine is most likely to already trust. A product with 40 recent, detailed reviews and a real 4.4 average is a far easier thing for an engine to recommend confidently than one with 6 reviews from two years ago, even if the underlying product is identical.
Where this fits in a bigger GEO plan
Comparison visibility is one piece of a broader e-commerce AI-visibility problem, and it's worth treating it that way rather than as an isolated content project. If you haven't audited crawler accessibility on your product catalog, start with AI visibility for e-commerce brands. If you're trying to get individual SKUs surfaced directly inside ChatGPT's shopping features, see how to get products recommended in ChatGPT Shopping. And if you're weighing whether this is a project you can budget for realistically, GEO pricing for e-commerce brands lays out real ranges instead of vague retainer talk.
What none of that replaces is the same discipline behind every AI-visibility fix we've covered in this publication: pick the highest-intent query, build the single clearest source for it, make the facts machine-readable, and repeat that for the next query once the engines have had time to re-crawl. Comparison queries just happen to sit closest to revenue, which is exactly why they're worth doing first.
How to tell if this is actually happening to you
Most DTC owners don't have a dashboard that says "AI engine answered your comparison query without a click," because no analytics platform ships that report by default. What you do have is a proxy: pull the Search Console query list for your comparison-style content, filter to queries with "vs," "or," and "better than," and look at the impressions-to-click ratio over the last six months. A page holding steady or growing impressions while clicks flatten or drop is the signature my client's humidifier post showed — the query volume didn't disappear, the click behind it did. Cross-reference against server logs or a bot-tracking tool for GPTBot, PerplexityBot, and ClaudeBot hits on that specific URL; a page getting crawled repeatedly by those agents while human clicks stay flat is strong evidence it's being read and cited, just not visited.
This is also where tracking AI referral traffic earns its keep — even a rough count of sessions arriving with a chatgpt.com or perplexity.ai referrer tells you whether the fixes below are moving anything, because the metric that used to prove a comparison page's value — organic clicks — is no longer the whole story.
What not to do
Don't respond to this by burying your product pages under scraped-together comparison content generated at volume with no real specs behind it — that's the fastest way to get flagged as unhelpful under Google's own helpful content guidance, and it doesn't fool an engine any better than it fools a shopper. Don't try to game the comparison by omitting a real competitor weakness in your favor either; engines cross-reference enough independent sources that a one-sided page just gets quietly excluded from the answer while a more balanced third-party page gets cited instead. The fix here is depth and honesty on a small number of pages, not volume on a large number of thin ones.
Questions people ask
The engine pulls specs, price, and reviewer opinions from whatever pages it can crawl and cite — often review sites, Reddit threads, and comparison articles rather than either brand's own product page — and gives a direct recommendation inside the chat window. No click is required to get an answer, which means neither brand's site gets the visit a search-engine comparison query used to send.
Not by paying for placement — there is no ad unit for this. A brand can influence it by making its own product page the clearest, most specific, most fact-dense source available: exact specs in structured data, honest limitations stated up front, and third-party reviews the engine can cross-reference. Engines favor pages that reduce their own work of verifying a claim.
Yes, carefully. A comparison page written in the brand's own voice, naming real competitors honestly including where a competitor wins, gives an AI engine a single well-structured source to cite instead of assembling an answer from five different third-party pages. It has to be genuinely fair or engines and readers both discount it; a page that only flatters itself reads as marketing and gets treated as marketing.
Want this done for you?
Want to know if AI is recommending you or your competitor? Get an AI visibility audit from Acromatico.
Get a free AI Visibility Audit →