AEO for affiliate sites: the citation and recovery playbook

An affiliate page gets cited in an AI answer when it contains something the answer can't get elsewhere: a test, a measurement, a named reviewer's verdict with reasons. It gets passed over when it restates the manufacturer's page with buy buttons attached. Since May 2026 that line is also a policy line. Google's spam policies now explicitly apply to AI Overviews and AI Mode, so thin affiliation costs AI visibility as well as rankings. For most affiliate sites the fix is to rebuild fewer pages properly, with real reviewers and real evidence, rather than to publish more.

Why affiliate content is exposed

"Best X for Y" and "X vs Y" are the queries affiliate sites live on, and AI Overviews increasingly answer them. Semrush tracked more than 10 million keywords through 2025 and found the share of AI Overview keywords with commercial intent rising from 8.15% to 18.57% between January and October, and transactional from 1.98% to 13.94% (Semrush AI Overviews study). Overviews started on informational queries. They no longer stay there.

The click cost is measured. Seer Interactive found organic CTR on informational queries with an AI Overview falling from 1.76% to 0.61%, and found brands cited in the Overview getting about 35% more organic clicks than uncited ones (Search Engine Land on Seer). Seer measured informational queries, not comparison queries, so treat it as direction rather than a forecast for your "best X" pages. The direction is clear enough: the Overview appears either way, and being one of its sources is the better side of the gap.

What changed in 2026 for affiliate sites

Two of these deserve a closer look.

The spam policy extension. On 15 May 2026 Google clarified that its spam policies apply to generative AI responses and added "attempting to manipulate generative AI responses in Google Search" to its definition of spam. The thin affiliation section was always aimed at pages that copy product descriptions and reviews from the merchant without adding value. It now covers your chances of being an AI Overview source, not only your blue link.

The August spam update. Glenn Gabe's case studies include a site built programmatically around Amazon product pages and affiliate links, with text the author assessed as largely AI-generated, that lost visibility for more than 14,000 queries. It's one example, not a census, but it is the pattern the thin affiliation and scaled content abuse policies describe.

What gets a review page cited

A named reviewer you can verify. "Reviewed by the editorial team" gives a reader nothing to check. A byline that links to a real bio page, with profiles that exist and describe the same person, does. Google's helpful content guidance asks exactly this: "Do bylines lead to further information about the author?" (Google). Ahrefs found brand web mentions correlating with AI Overview brand visibility at 0.664, against 0.218 for backlinks, across 75,000 brands (Ahrefs). That study is about brands and correlation, not proof that a reviewer's name causes citations, but a reviewer with a public presence is how mentions can lead back to your pages.

Evidence of actual use. Google's reviews guidance asks for "visuals, audio, or other links of your own experience," quantitative measurements, and benefits and drawbacks "based on your own original research." Content written without using the product restates the spec sheet because there's nothing else to say.

A comparison table. Consistent columns (price, the spec that matters, who it's for, the main drawback) are easy for a reader to scan and for an engine to lift. On "X vs Y" pages the table is often the most quotable thing you have.

The verdict near the top. If the recommendation only appears in the conclusion, a reader skimming the top, and an engine pulling a passage from it, miss it.

Honest disclosure. Covered in the checklist below, because it's where most sites get the details wrong.

Rebuild checklist

Work through it in order. Diagnosis comes first because a spam-update hit and a core-update decline need different fixes. The full diagnostic is in clicks dropped but rankings didn't; the affiliate-specific steps are here.

1. Diagnose before you change anything

Match the drop date to the Search Status Dashboard. 2026 so far: March spam (24–25 Mar), March core (27 Mar–8 Apr), May core (21 May–2 Jun), June spam (24–26 Jun), August spam (18–21 Aug), September spam (from 24 Sep, still rolling out on 1 Oct).
Decide: spam or core? A drop during a spam update points at a policy violation; check thin affiliation and scaled content abuse first. A drop during a core update is a quality reassessment.
Check whether you lost rankings or only clicks. If position held while clicks fell, an AI Overview is the likely cause, and the generative AI report in Search Console can confirm it.
Check Manual actions, Security issues and the Page indexing report in Search Console.
Pause any pipeline that auto-publishes AI-generated pages without substantive human review. Mass-generated pages are what the scaled content abuse policy targets.
Build an audit sheet of your top 50 pages by historical traffic. Columns: clicks and links (before and after), named reviewer (Y/N), first-hand evidence (Y/N), last substantive update, disclosure placement, affiliate link markup.

2. Reviewer infrastructure

Pick one to three named specialists, not ten thin bylines. A narrow site is better served by one person who owns a category.
Give each a real bio page: background, why they can judge these products, a real photo, links to profiles that exist.
Add Person markup to the bio page and reference it from each review's author (with a url to the bio). This helps search engines attribute authorship correctly. It is not a ranking factor in itself, and Google's AI features documentation says no special markup is needed for AI Overviews.
Never invent reviewer personas. Stock photos with made-up credentials are worse than no byline.

3. Per page, for the pages worth rescuing

Re-test or re-research the products. Update prices, current models and alternatives. Replace discontinued picks.
Change the date only when the content changed. Google warns against "changing the date of pages to make them seem fresh when the content has not substantially changed."
Add original evidence: your own photos, measurements, screenshots, the problem you hit in week two.
Put the verdict, and who it's for, in the opening paragraph.
Build or fix the comparison table.
Answer real buyer questions in a visible section ("is it worth it," "X vs Y for my use case," "what's the return policy"). FAQPage markup is optional and has produced no rich result since 7 May 2026, so don't add it for Google's sake.
Keep Review markup honest. The rating must match the page, and since 24 July 2026 Google's review snippet guidelines say "Don't include fake or undisclosed incentivized reviews on your page or in your structured data markup."
Qualify affiliate links with rel="sponsored". Google prefers it for paid links and still accepts nofollow.
Make the disclosure clear and close to the recommendation. The FTC says to disclose "clearly and conspicuously" and that "the closer the disclosure is to your recommendation, the better" (FTC endorsement guides FAQ). A footer or a separate disclosure page doesn't meet that. One line near the top and next to the links works.

4. Pages you can't save

Improve first, consolidate second, delete last. Google calls deleting content "a last resort" (core updates guidance). Merge near-duplicates into the strongest page with a 301. Remove only pages with no traffic, no links and no realistic path to being useful.
Return 404 or 410 for removed pages. Google treats them the same: "All 4xx errors, except 429, are treated the same" (Google).
Don't rely on the Removals tool. It only hides a URL, and "a successful request lasts only about six months" (Google). Permanent removal comes from the 404/410 or a noindex.
Take removed URLs out of your sitemap.

5. Measure

Record a baseline before you start: clicks, impressions and position per page, plus AI Overview impressions from the generative AI report.
Sample ChatGPT and Perplexity on a fixed set of 10–20 queries at the same interval, and note who gets cited.
Check server logs for the retrieval crawlers: Googlebot, OAI-SearchBot, Claude-SearchBot and PerplexityBot should be fetching the rebuilt pages. Google-Extended is a robots.txt token with no user agent of its own, so it never appears in logs (Google). GPTBot and ClaudeBot are training crawlers and don't decide citation (Anthropic).
Change one group of pages at a time, so you can tell what worked.

Free tool · No signup

Free AI Crawler Checker

Paste a domain → see which AI crawlers your robots.txt allows and which it blocks, with training and retrieval bots told apart.

Check your crawlers →

Rebuild or publish more?

Rebuild, for most sites. A 2023 Ahrefs study of about 14 billion pages found 96.55% get no organic traffic from Google (Ahrefs). Thirty new articles on a site Google is currently reassessing will mostly reach nobody.

A page that already ranks has cleared the hardest stage. In the C-SEO Bench study of product recommendations, moving up in retrieval rank was roughly 7.7 times as effective as the best content-rewriting method (2.77 against 0.36 in the retail task, Puerto et al., 2025). Improving a page that's already retrieved works on something that passed that gate.

New writing wins only where existing pages have nothing to build on: no traffic, no links, no reviewer who used the product.

What to expect

Google says it "could take several months" for its systems to confirm a site-wide improvement after a core update, and that some changes show up sooner, including in smaller core updates. For spam updates, Glenn Gabe's case studies quote Google saying recovery "can take months after making significant changes," and mention one site that recovered about five months after cleanup. ChatGPT and Perplexity fetch pages through their own retrieval, so a rebuilt page can start appearing there before Google has re-evaluated the site, or not at all; that's why the checklist measures them separately. Nobody publishes reliable "percent recovered by week N" data for affiliate sites, so this page doesn't give one.

The ceiling matters as much as the timing. Even a fully rebuilt page ranks into a results page where Ahrefs measured the top result's CTR 58% lower when an AI Overview is present (Ahrefs, Feb 2026). Being the cited source is the part you can still win.

When to stop rescuing

Sometimes a fresh, focused site is the better business decision. Take it seriously if most pages are anonymous, untested and have no history of traffic; if you aren't the right person to review the products; or if the rebuild would take months you can't fund. Make that call after the audit, not in the first week after a drop.

FAQ

Can I use AI to draft affiliate reviews?

Google's concern is the absence of value, not the tool. A draft that you then fill with your own testing notes, photos, measurements and drawbacks can be a strong review. A draft published with light polishing is the pattern the scaled content abuse and thin affiliation policies describe, and since May 2026 those policies apply to AI Overviews and AI Mode as well.

Should I opt my affiliate site out of AI Overviews?

Search Console's generative AI control, live for all sites since 31 August 2026, removes your pages from AI Overviews and AI Mode as links and as grounding. Google says rankings are unaffected. It won't return clicks, because the Overview still appears for the query using other sources. For most affiliate sites, being one of the cited sources is worth more than being absent.

Do I still need Review markup if AI features don't require structured data?

For review stars in regular results, yes: Review snippets remain a supported rich result, as long as the rating matches the page and the reviews aren't fake or undisclosed incentivized. For AI Overviews and AI Mode, Google says no special markup is needed, so the markup won't compensate for a thin review.

Start with one page

Pick the affiliate page that lost the most clicks. Check its drop date against the dashboard, then check whether position held. If it held, the work is evidence and a verdict up top. If it fell during a spam update, read the thin affiliation policy against that page line by line before you edit anything. For the markup details, see Review schema for affiliate sites; for the pre-publication list, the AEO content checklist.

Sources cited in this piece

Updated October 1, 2026: merged our affiliate recovery checklist into this page with corrected technical details (404 vs 410, the Removals tool, crawler names, FTC disclosure), and added Google's 2026 spam policy, review snippet and AI optimization changes and the August and September spam updates.