An affiliate or content site earns money when a reader clicks through to a merchant and converts. That model assumed the reader found your page in search results, read your comparison, and trusted your link. A growing share of buyers now start somewhere else: they ask ChatGPT, Gemini, Perplexity, or Claude “which [product] should I buy,” and the assistant answers with a short recommendation and a handful of links. Some of those links go to your page. More often, they go straight to the merchant — or nowhere.
Why this segment is different
If you run an affiliate or content site, “AI visibility” is not a separate marketing channel to add to the spreadsheet. It is a threat to the exact moment your revenue happens: the click. This post covers what actually happens to affiliate pages inside AI answers, how to measure it honestly, and the disclosure questions that come with it — all grounded in what is publicly documented today, with the sources named.
When a buyer asks an assistant a money question — “is [product X] worth it,” “best [category] under $200” — one of three things happens to your page:
- The assistant cites your page and links to it. The reader can click through to your comparison, and the affiliate journey continues as usual.
- The assistant uses your information but names the merchant, not you. The answer says “Product X is the best budget option” and links the merchant’s site. Your research did the work; you receive nothing.
- The assistant does not draw on your page at all. Another site, a community thread, or nothing gets referenced.
Each outcome needs a different response, and you cannot respond without knowing which one is happening — that is the measurement problem, and what the rest of the post covers.
There is no affiliate-specific citation data published yet. No assistant has documented how it selects product-category sources, and no independent study has measured citation rates for affiliate sites specifically. What exists is category trackers produced by AI-visibility SaaS tools, and we do not adopt vendor pipeline numbers. So this segment has to be measured the same way we measure everything else: with your own logs, labeled observational, not causal.
The click side of the problem is better documented. AI Overviews are associated with sharply lower click-through to position one on the queries where they appear: Ahrefs found position-one CTR was 58% lower in December 2025 than in December 2023 across 300,000 desktop keywords (Ahrefs, February 2026, observational: ahrefs.com/blog/ai-overviews-reduce-clicks-update). We break the published studies down in Ahrefs says −58%, Seer says −61%: what the AI Overviews CTR numbers actually mean.
For a brand that converts on its own site, a traffic dip is a traffic dip. For an affiliate site, the click is the revenue. An answer that names the merchant without your link is not a lost impression — it is a lost commission, with no impression to recover.
The disclosure question
Disclosure rules were written for human readers, but affiliate pages now have a second, uncontrollable distribution channel: the assistant’s summary. When an assistant rewrites your comparison, you cannot control which sentence of your page — disclosure included — makes it into the answer. The obligation, therefore, stays entirely on your page, and it does not change just because assistants consume it.
What the rules actually require, today:
- US (FTC): Material connections must be disclosed “clearly and conspicuously.” A commission arrangement is the textbook material connection — 16 CFR §255.5 defines it explicitly, including payment and free or discounted products (eCFR, current as of October 28, 2026). The rule text sets the standard; the FTC’s staff guidance fills in placement practice.
- UK (ASA/CAP): Marketing communications must be “obviously identifiable” as such, and the CAP Code applies to affiliate content in full — a publisher who earns commission on a link is acting as a secondary advertiser (ASA online advice, updated March 22, 2023).
- EU: There is no single EU-wide affiliate-disclosure rule; requirements come from national consumer-protection and marketing law, enforced by national authorities. Check the rules of the markets you publish in before relying on a US-style footer.
Three practical consequences for an AI-first strategy:
- Keep the disclosure in static page text, near the money links. Buried in a terms page or footer it fails the “clear and conspicuous” standard for human readers — and it is also the least likely text to survive into a summarized answer.
- Do not assume the assistant’s answer carries your disclosure. If the assistant names the merchant without your page, no one reading the answer has seen your disclosure; it is an argument for putting the disclosure where a crawler reads it first, not last.
- Disclosure is not the same as quality. Disclosure keeps you legal; the content is what gets you cited.
How to measure the three outcomes
The method is the same baseline test we use for every brand and segment, adapted to affiliate economics: a fixed list of your revenue-driving questions, asked on every major surface, logged response by response, and scored into the three outcomes from the first section. It is manual, it is small-sample, and it is the only measurement that can tell you which outcome a specific question produced. The full procedure, including how questions are selected, is in our methodology.
Step 1 — Build the money-question list. Write down the 20–30 questions that drive the site’s revenue: the “best X under Y” queries, the head-to-head comparisons (“X vs Y”), and the “is X worth it” pages that earn the most commissions. Rank them by affiliate revenue where the network dashboard allows it. This list is your exposure map — the answers where your site should appear.
Step 2 — Run the monthly baseline test. Pick the 10–20 highest-revenue questions from that list and ask each of them, every month, in ChatGPT, Gemini, Perplexity, and Google AI Overviews. Log every response: date asked, exact question, whether your page was cited or your name/brand appeared, the page URL cited, whether a link back to your site was present, what the answer said about the merchant, and which sources were cited instead.
Step 3 — Score each question into one of the three outcomes. Cited with link; used without naming; not drawn on. Track the count per outcome month over month, per surface.
Step 4 — Read the log, not the aggregate. The counts move slowly; the reasons move fast. A question where the assistant consistently names the merchant and links Amazon is a different situation from one where it cites a Reddit thread instead of you, and each points at a different fix. And read it with the AI-Overviews caveat in mind: on Google, a query that gets an Overview is also a query where position-one clicks are structurally lower, so a flat “cited with link” count on Google can coexist with a falling click count.
One clarification that matters here: AI visibility is separate from Google rank. A page can rank number two in Google and never be cited by an assistant, or rank on page three and be the source an assistant reaches for, because assistants retrieve and cite from their own crawled index rather than Google’s rankings. If your monthly report has no row for “cited by assistants,” you do not yet have an AI-visibility measurement — you have a traffic report with a new referrer.
Measuring click impact
Citations are the leading indicator; commissions are the number. The click side has four data sources, listed in order of reliability, and each answers a different question: whether the click happened, where it came from, whether Google’s own AI surfaces drew on the page, and whether the two agree. None is complete; together they bracket the true number.
1. Your affiliate network dashboards. Per-link conversions and earnings, monthly, per top page. This is the only number that is not an estimate. Sample sizes per question will be small — often single digits per month — so report direction over three months, not a weekly spike.
2. GA4 AI referrers. Assistant traffic arrives in GA4 as ordinary referral rows from the surface’s domain (chatgpt.com, perplexity.ai, claude.ai, gemini.google.com). Two caveats before you build a report on it: assistants frequently present your link without your UTM parameters, and some referrals arrive with the referrer stripped, landing in direct. The GA4 AI-referral number is a floor, not a census. Setup, segmentation, and the caveats in full: Measure AI search traffic in 2026: GA4, GSC and Bing.
3. GSC’s generative AI report. Google’s report (launched June 3, 2026, all websites worldwide since August 31, 2026) shows impressions of your site inside AI Overviews and AI Mode — no clicks, no citations. It tells you where Google’s own AI surfaces draw on you; it says nothing about ChatGPT or Perplexity, and it says nothing about what happened after the impression.
4. The baseline log, joined to the above. The join that does not exist anywhere else: for each logged question, do the assistant-cited clicks show up in GA4 referrers and network conversions? Over a quarter you will have too few data points for statistics, but enough for the honest version of the answer: “when the assistant cites us, a small number of real affiliate clicks follow.”
Two attribution warnings, stated plainly. First, last-click attribution to organic search overstates the AI-driven decline: a buyer who asked ChatGPT, clicked the merchant link, and bought there never appears in your data at all. Second, the reverse error is just as common — assuming every perplexity.ai referral is a win, when the reader may have clicked once, seen the merchant’s price, and converted on the merchant’s site without your link. Both errors point the same way: treat the numbers as a floor and a ceiling around an unknown middle, and make content decisions on the log, not the totals.
What actually moves outcomes
There is no published list of the factors an assistant weighs when choosing a product-category source, and no vendor will hand you one. What is available instead is a smaller, defensible set: page fundamentals that make a page usable at all, claims sharp enough to be the cited answer, and the share of citations community and third-party sources carry — each backed by the sources named below — plus freshness, where we state plainly what is documented and what is judgment.
- Citable page fundamentals. Extractable claims, structured data, clean crawl access, and a clear answer to the question on the page. The full checklist: What makes content AI-citable. This is where outcome 3 (not drawn on) usually starts: the page is blocked, thin, or unfocused.
- Citable claims, not citable pages. A page is cited when a specific claim on it is the best available answer to a specific question — specific, sourced, and fresh. That is the AEO/Geo core, in one page: AEO/Geo basics in one page. For affiliate pages this usually means turning a product roundup into a set of sharp, checkable claims: price, warranty, measured specs, failure modes — with the affiliate disclosure intact.
- Community and third-party sources carry a large share of citations. Ahrefs’ monthly most-cited-domains study, September 2026 edition, finds reddit.com at 16.8% of top-source mention share in ChatGPT, with other community and reputation domains high in the list (Ahrefs, updated September 2, 2026: ahrefs.com/blog/most-cited-domains-in-chatgpt; analyzed in Why AI answers triangulate). An affiliate site’s realistic path into outcome 1 is being one of the few first-party sources in the conversation — a hands-on review with original numbers — rather than another aggregation of the same spec sheet.
- Freshness is judgment, not a documented weight. Price, availability, and feature claims are time-boxed facts, and a review whose numbers have gone stale is a source an answer is worse off drawing on — so keep pages dated and updated, and treat that as a hedge rather than a lever. No assistant has published a weight for freshness, and none of the published studies we cover isolate it. Like everything else here: observational, not causal.
The honest summary: none of this is a ranking factor you can optimize for a guaranteed outcome. It is the difference between a page an assistant can use and one it cannot — and, for this segment, that difference is the difference between commission and nothing.
Start with the $99 audit
If you run an affiliate or content site and cannot say which of the three outcomes your money pages are in, that is exactly the gap the AI Visibility Audit closes: a one-off review of your revenue questions across the major assistant surfaces, delivered as a PDF you can act on, for $99, with no subscription attached.
How it works: email hello@aecited.com with your site and a few of your key money questions. Within one business day we reply with a set of 30–50 draft buyer questions for you to review, plus a payment link. After you confirm the questions, the PDF is ready within three business days. It covers the money questions, the three outcomes per answer, the click path, and the disclosure review.