Hi everyone, this is Neo.

Today’s post starts with something from this week’s SEO Pulse over at Search Engine Journal — a piece of news I think deserves a much closer look for everyone running an independent online store.

Here’s the headline: AI tools recommend your brand, but hand the citation to someone else.

Translation: you’ve worked your brand into AI’s answers — ChatGPT names your brand when someone asks which squat-proof leggings are worth buying — but the source link sitting right under that answer points to a third-party publisher like Good Housekeeping or Verywell Fit, or maybe even your competitor.

Your brand gets the shout-out. Someone else gets the traffic. This isn’t an isolated quirk anymore; it’s turning into a clear industry pattern. Today I want to unpack this “recommendations vs. citations” split with the freshest data available, then talk about what independent sellers should actually do about it.

The Numbers That Should Send a Chill Down Your Spine

Let’s start with the star of this week’s news cycle: Shero Commerce (a Shopify agency that specializes in SEO and AI services) published a study called Duplicate Content and AI Citations.

Here’s what they did — it’s pretty ruthless. They ran real buying questions across Google AI Mode, ChatGPT, and Perplexity in 60 ecommerce categories, collected all 1,851 sources those AI answers cited, and checked each one: how many of them were the brand’s own website?

The answer: just 2.8%.

In other words, when AI answers “what should I buy” questions, fewer than 3 out of every 100 cited sources are brand-owned pages. The rest are media, review sites, marketplaces — third parties.

And here’s the gut-punch follow-up: when AI named a brand by name in its answer (they caught 159 such recommendations across all platforms), that brand’s own website got the citation only 31% of the time.

Real-world example: for a squat-proof leggings query, both ChatGPT and Perplexity recommended Gymshark, Alo, and Beyond Yoga by name — then cited third-party sites as the sources.

Google AI Mode was even worse: brands were cited or recommended in just 9.5% of the relevant store checks across 60 categories, and in about a third of categories, none of the tested brands showed up at all.

And this isn’t one agency’s fluke. Victorious (a US SEO agency) confirmed the pattern at a much bigger scale in its Q2 2026 Quarterly Search Report: 175 brands, 8 AI platforms, 5 verticals, and 49,391 citations across 5,830 AI-generated answers — of which 99.99% pointed to third-party websites. Only 4 brands in the entire study ever received a citation to their own domain. ChatGPT and Google AI Mode each cited a brand’s own site once or twice across more than 25,000 combined citations. Gemini and Google AI Overviews never did it once.

Then there’s the gap between recognition and mentions: AI accurately described 96% of the tested brands when asked about them directly (ask it “what does this company do” and it nails it), but when asked the category research questions real buyers actually ask (“who’s best in this category”), 89% of those brands never appeared at all.

A German B2B study tells the same story: across 62 B2B brands, average AI Share of Voice was just 6.6%; only 5% of AI citations pointed to a brand’s own website — the other 95% went to Reddit, Forbes, Wikipedia, YouTube — and 69% of AI answers cited a brand without recommending it.

Three independent studies, one conclusion: AI knows you. It may even mention you. But it won’t give you the citation — the actual click-through entry point — unless something changes.

Why Recommendations and Citations Are Coming Apart

To fix a problem, you have to understand the machinery behind it. As I see it, there are three layers at work.

First: AI’s “answer layer” and “retrieval layer” are separate processes.

When AI generates an answer, it first retrieves relevant pages from the web, then synthesizes them into a response. Along the way, “who to recommend” and “who to cite” can be driven by two different logics — recommendation tracks a brand’s overall footprint in the information ecosystem, while citation tracks which specific page is the best source material for this particular question.

Aleyda Solís summed it up perfectly on LinkedIn: “Your brand can get recommended in the answer… while the citation (and the click) goes to a magazine or a marketplace.”

Second: your content has been copy-pasted into someone else’s content.

Shero’s study surfaced a detail worth pausing on: across the 1,000 Shopify stores they analyzed, 20% of product descriptions appeared word-for-word on another domain.

This is what happens when brands syndicate the same copy everywhere — Amazon, AliExpress, distributors, wholesale partners. AI crawls the web, finds the same description on a hundred domains, and naturally treats the most authoritative or earliest-seen page as the source. Which is usually not your website.

You wrote the copy. Someone else gets the citation. Classic.

Third: media and review sites live on AI’s “trust list.”

59% of AI citations went to publishers like Good Housekeeping, Verywell Fit, and Reviewed.com. That’s not hard to understand: these sites have accumulated decades of backlinks and authority in the old search era, and when AI weighs “which source is most trustworthy,” it leans their way. Your store might have better content — but on the trust axis, you’re fighting uphill.

Behind All This: A Fight Over What Content Is Worth

Let’s zoom out for a second. This week’s SEO Pulse also carried a story that’s basically the mirror image of this one — the latest development in the Penske Media vs. Google case.

Penske owns Rolling Stone, Variety, and other major publications. Their antitrust suit against Google alleges that Google uses its search monopoly to force publishers to hand over content for AI training and AI Overviews — without paying a dime.

At Tuesday’s hearing, Judge Amit Mehta pushed back on Google’s characterization of AI Overviews as a “product improvement.” Jason Kint (CEO of Digital Content Next, who was in the courtroom) reported that Mehta described the arrangement as “seems really unfair” — Google’s AI products are being built “on the backs of the publishers.”

Mehta also flagged the contradiction at the heart of it: other AI companies, like OpenAI, are paying millions for content licensing deals, while Google can consume the entire internet for free thanks to its search monopoly.

Read these two stories side by side and they’re really the same story: in the AI search era, content producers — publishers and independent sites alike — are losing control over how their content turns into money. Your content feeds the AI, the AI answers questions with your content, and the traffic doesn’t necessarily come back to you.

Don’t Rush Into “Special GEO” Work: Listen to Google Itself

So if AI’s recommendations and citations are this tangled, should we all sprint off and build a “special AI optimization” (GEO) playbook?

Funny you should ask — John Mueller addressed exactly this on Bluesky this week. An SEO asked whether there are industries where GEO doesn’t matter yet. Mueller’s answer: “From our POV there’s nothing really special you need to do for generative AI responses in search.”

Note the qualifier: “from our POV.” Mueller is talking about Google’s own AI search — AI Overviews and AI Mode. He’s not speaking for ChatGPT, Perplexity, or any other engine, and those systems retrieve and select sources differently. What Google doesn’t need isn’t automatically what everyone else doesn’t need.

But I think the core of Mueller’s point is worth taking seriously: don’t turn GEO into a mystical checklist. The real work still lives in the boring, durable fundamentals — original content, third-party brand validation, pages that crawl cleanly. Get those right and you won’t be caught flat-footed no matter how AI search evolves.

Neo’s Take: What Independent Sellers Should Actually Do

Enough analysis. Here are five practical moves, all aimed at one goal: winning back the citations being stolen from you.

1. Track “mentions, citations, and clicks” separately — stop watching a single number.

This is the most overlooked move. Most store owners measure AI visibility by one question: “Did my brand show up in AI answers?” By now you know better: showing up ≠ being cited ≠ getting traffic.

Steal Victorious’s framework and track three metrics independently:

• Mention rate: does AI name your brand?
• Citation rate: does AI link to your website as a source?
• Click rate: do users actually land on your site?

Only when you track all three can you tell whether your AI marketing is real or cosmetic. If your brand gets named constantly but never cited, the problem is content-level, not awareness-level.

2. De-duplicate your product copy. Now.

Shero’s finding that 20% of product descriptions are copied verbatim on other domains is the most avoidable way to hand away citations.

Here’s my advice: rewrite the descriptions you syndicate to every channel (Amazon, distributors, wholesale platforms) — at minimum for your highest-intent product and category pages. Make them original, answer-extractable copy. When AI scans the web and finds this copy exists only on your site and everyone else is copying you, the citation has nowhere else to go.

3. Stop pouring 100% of budget into your own site. Build third-party validation.

Victorious found the signals most strongly tied to AI mentions were referring domains (r = 0.49) and third-party web mentions (r = 0.45) — neither of which is a metric on your own site.

The German B2B study put it bluntly: build authority in the places AI already trusts (Reddit, industry directories, trade media, Wikipedia) — a few mentions there beat hundreds of links on small irrelevant sites. And content published under a named human expert gets cited more often than anonymous corporate posts.

So carve out part of your marketing budget for: media interviews, product reviews, directory listings, genuine Reddit/forum discussions, YouTube review videos. Make AI keep seeing you on other people’s turf.

4. Rewrite your highest-intent pages as extractable answers.

Google AI Overviews and Perplexity pull from the live web, and they love content shaped as direct answers. Rebuild your top product and category pages so they answer buyer questions head-on — clear specs, honest comparisons, direct pros and cons — so that when AI crawls your page, the answer is right there to lift.

5. Watch who gets the citations — and turn it into a PR hit list.

Every week, throw your category’s real buying questions at ChatGPT, Perplexity, and Google AI Mode. Record who gets recommended and who gets cited. You’ll spot a pattern fast: the third-party sites that keep getting cited are exactly the media resources you should be pitching. If Good Housekeeping keeps getting the citation today, your job is to get Good Housekeeping to write about you tomorrow.

Wrapping Up

“Recommendations vs. citations” is a phrase I expect to become a buzzword in the independent site SEO world this year. It captures the harshest reality of the AI search era: content production and traffic distribution have fully decoupled.

But look at it from the other side — this is precisely the opportunity. Most of your competitors are still obsessing over “how do I get AI to mention my brand,” while you now know that being mentioned is worth less than being cited, and being cited is worth less than being clicked.

Get all three layers working, and you’ve genuinely earned a seat at the AI traffic table.

This is Neo. See you in the next one.