Getting mentioned on other websites

Andy Bright

Director of Product and Technology

When an AI assistant recommends a product, much of what it says comes from what other people have written. Review sites, newspaper lists, comparison articles, forum threads and videos all feed into its answer. For most businesses, what other websites say about them can affect AI answers more than anything on their own site.

This chapter explains the evidence for that, how to find the websites that matter in your market, and how to earn mentions on each kind of site without breaking platform rules or the law.

Why other websites matter so much

Marketers divide the places a business appears into three kinds. Owned media is what you publish yourself, such as your website. Earned media is what independent people publish about you, such as reviews, news articles and lists of recommendations. Paid media is advertising.

AI answers lean towards earned media. A 2025 study from the University of Toronto⁠(opens in a new tab) asked a search version of GPT-4o, an OpenAI model available to developers, to rank cars, electronics and software in the US and Canada. It compared the model's sources with Google's top results. Between 69% and 92% of the AI's sources were earned media, against between 32% and 54% of Google's. The authors describe "a systematic and overwhelming bias towards Earned media".

Ahrefs, which sells SEO tools, looked at 75,000 brands in December 2025 to see what the brands that AI assistants mention most often have in common (Ahrefs⁠(opens in a new tab)). In this chapter a mention means your business's name in the text of another website, video or forum post, with or without a link. Ahrefs measured each possible factor with a correlation, a number between −1 and 1 that shows how closely two things rise and fall together. A correlation near 1 means they rise together closely, one near 0 means there's no relationship, and a negative one means one falls as the other rises.

Mentions on other websites correlated strongly "across every AI platform we studied", which covered ChatGPT, Google's AI Mode and AI Overviews. The size of a brand's own website barely mattered. Backlinks are links from other websites to yours, which search engines use as one signal when they rank pages. An earlier Ahrefs study of Google's AI Overviews, in May 2025⁠(opens in a new tab), found that backlinks correlated much more weakly (0.22) than mentions did (0.66).

Correlation doesn't prove cause. Big, well-known brands have more of everything, including mentions, videos and AI recommendations, so these studies can't show that adding mentions will make assistants recommend you. But they all point the same way as the Toronto study. In these studies, mentions by other people went with AI mentions far more closely than the number of pages on a brand's own site.

Google warns against trying to fake this. Its guide to optimising for generative AI⁠(opens in a new tab) says "seeking inauthentic 'mentions' across the web isn't as helpful as it might seem". This chapter covers ways to get real people to write about you, in places your customers trust.

Find the websites that matter in your market

Which websites assistants rely on depends on the assistant, the kind of question and the industry, and it changes from month to month, as the chapter on each assistant shows. To find the websites that matter for your customers' questions, run your own checks.

If you did the exercise at the end of the first chapter, you already have a list of the websites the assistants linked to. Group them by kind, and note whether each one mentions you. Fernhill Coffee's list looked like this:

The websites that appear again and again, and that mention competitors but not you, are where to start.

Reviews

Customer reviews matter in two ways. AI assistants read review sites when they compare businesses, and some summarise reviews directly. OpenAI says the review summaries in ChatGPT's shopping results are "based on reviews from public websites" (Shopping with ChatGPT search⁠(opens in a new tab)). A complaint that appears in many reviews is likely to appear in the summary too.

Each review site has its own rules

Review sites set different rules about asking for reviews and rewarding people for writing them. An incentivised review is one where the reviewer gets something in return, such as a discount or a gift card. A scheme that's allowed on one site can break the rules of another.

• G2, a review site for business software, allows incentives only if they're disclosed to G2, don't depend on the review being positive, and are worth no more than $100. It labels these reviews "Incentivized". It also bans "collecting reviews by any method that segments out potentially negative reviews" (G2 community guidelines⁠(opens in a new tab)).
• Trustpilot doesn't allow them at all. Its guidelines for businesses⁠(opens in a new tab) say "We don't allow incentivized reviews on Trustpilot", and businesses can't pick and choose which customers to invite.

Read the rules of each review site you use before you offer any reward or start a review campaign.

The law on reviews

Fake reviews are illegal in the UK and the US.

In the UK, the Competition and Markets Authority (CMA) published guidance on fake reviews⁠(opens in a new tab) in April 2025. It explains a ban on fake reviews, on "concealed incentivised reviews", meaning reviews that hide the fact that the writer was rewarded, and on publishing reviews in a misleading way. The guidance gives "encouraging just those who are satisfied to leave reviews" as an example of publishing reviews in a misleading way.

Businesses that publish reviews, including on their own website, must also take reasonable and proportionate steps to prevent and remove fake ones. The CMA's short guide for businesses⁠(opens in a new tab) says this includes a published policy that bans fake reviews, and processes to find, investigate and remove them. The CMA checked more than 100 business websites after the ban came in and found that 54 could be failing to comply. In March 2026 it opened investigations into five businesses over their online reviews (CMA case page⁠(opens in a new tab)).

In the US, the Federal Trade Commission (FTC) has enforced a rule on reviews and testimonials⁠(opens in a new tab) since October 2024. It bans fake reviews, "such as AI-generated fake reviews", and rewards that depend on the review being positive or negative. It also bans reviews from staff or managers that don't disclose the connection, review sites a business controls but presents as independent, and suppressing negative reviews through threats. The FTC can seek civil penalties against businesses that knowingly break it. Asking only happy customers for reviews isn't banned by the rule itself, but the FTC says⁠(opens in a new tab) it "could violate the FTC Act".

Other countries have similar laws. Check the rules in each country where you sell.

A review routine that follows the rules

1. Ask every customer, at the same point, whatever their experience. Trustpilot's guidelines specifically ban sending invitations only at a stage that happy customers reach. Asking only after a second order would be one example. Fernhill asks every subscriber a week after their first delivery, and asks again when anyone cancels.
2. Don't filter who gets asked, and don't make any reward depend on what people write.
3. Make it easy, with a direct link to the review page.
4. Reply to negative reviews publicly, calmly and with specifics. Customers read replies, and assistants can too.
5. Fix the problems that keep coming up. If ten reviews say deliveries to Ireland are slow, an AI summary is likely to say it too, until the reviews change.

Reddit and other forums

People ask forums the same questions they ask AI assistants, and the answers are often frank. Some assistants cite forums often, especially Google's AI Mode, while ChatGPT's use of Reddit has swung sharply during 2026, as the chapter on each assistant explains.

Reddit's rules⁠(opens in a new tab) ask people to "participate authentically in communities where you have a personal interest, and do not spam or engage in disruptive behaviors (including content manipulation)". Its informal guide to good behaviour, called Reddiquette⁠(opens in a new tab), suggests that "only 1 out of every 10 of your submissions should be your own content". Each community on Reddit, called a subreddit, sets its own rules too, and many ban self-promotion entirely.

If you take part:

• Use an account that clearly belongs to you or your business, and say who you are whenever your own product comes up.
• Answer the question that was asked, even when the best answer is a competitor or a cheaper option.
• Respond helpfully to threads about problems with your product, and fix the problem. That changes what the thread says.
• Never use fake accounts, bought upvotes (paid votes that push a post higher) or services that "seed" mentions of your business. They break Reddit's rules, and in the US the FTC treats endorsements that hide a connection to the business as potentially deceptive (FTC⁠(opens in a new tab)).

The same approach works on other forums, such as Quora, Mumsnet, Stack Overflow, Discord groups and industry forums.

Wikipedia

Wikipedia is one of the websites AI assistants cite most often, although ChatGPT cited it much less in the US in late September 2026 than in March (Ahrefs⁠(opens in a new tab)). Its rules on writing about your own business are the strictest in this chapter.

Wikipedia only has articles about businesses that independent sources have written about in depth. Its notability guideline for organisations⁠(opens in a new tab) requires significant coverage in several independent, reliable sources, and says press releases, news of a company raising money, and brief mentions don't count.

You're also strongly discouraged from writing your own article. New articles from people with a conflict of interest are meant to go through a review process called Articles for Creation. Wikipedia's conflict of interest guideline⁠(opens in a new tab) says people with a connection to a subject "are strongly discouraged from editing affected articles directly, and can propose changes on article talk pages instead". Anyone paid to edit must disclose who is paying them.

So the route to a Wikipedia article is to earn independent coverage first, and let volunteer editors decide. If an article about your business exists and is wrong, post the correction on the article's Talk tab, at the top of the page. Say who you are, and link to independent sources that support it. Wikidata, the database of facts behind Wikipedia, has its own rules, covered in the chapter on describing your business.

Lists of recommendations and roundups

When someone asks an assistant for "the best coffee subscription", the answer often draws on articles that list and compare the options, such as "The 10 best coffee subscriptions in the UK". These are called roundups or "best of" lists.

They're among the pages ChatGPT cites most. Ahrefs looked at the answers to 750 "best X" prompts it wrote, such as "best home treadmill", and found⁠(opens in a new tab) that "best X" lists made up 43.8% of the pages ChatGPT cited. Its early analysis also found that brands placed high on those lists seemed more likely to be recommended, though Ahrefs says it is still studying this and hesitates "to give a definitive conclusion".

To get into the roundups that matter:

1. Find the lists that assistants cite for your customers' questions, using the checks above.
2. Find out who writes each one, and whether and when it's updated.
3. Send the writer what they need to judge you fairly, such as accurate facts, prices and a sample or trial. Fernhill sends a bag of each coffee and its plan details.
4. If a list already includes you but gets something wrong, ask for a correction with the facts.
5. Don't pay for a place without it being disclosed. If you pay for inclusion, or the publisher earns a commission from your sales, that's advertising:
   • In the US, the FTC requires such relationships to be disclosed.
   • In the UK, the Advertising Standards Authority's rules require ads to be clearly identifiable as ads, as its guide for influencers⁠(opens in a new tab) explains.
   • Google's spam policies⁠(opens in a new tab) expect paid links to carry rel="sponsored" or rel="nofollow", two short labels in the link's code. The publisher adds these, so ask them to.

Lists on your own website that rank your own business first are a different matter, covered in the chapter on what to avoid.

Press coverage and original research

Give journalists, bloggers and other businesses something new to write about. Original data works especially well, because a journalist who quotes your figures names your business as the source.

Fernhill publishes a short report each autumn from its September customer survey, with a headline figure that journalists can quote, such as how much the average UK subscriber spends on coffee each month. A good research piece:

• answers a question people in your industry actually ask
• explains how the data was collected, and from how many people
• leads with one clear figure that's easy to quote
• is published as an ordinary web page with the main findings in text, not only as a PDF, because Microsoft advises⁠(opens in a new tab) against keeping core information only in PDFs
• is repeated each year, so people come to expect it and link to the latest edition

Expert comment works as well. Journalists writing about coffee prices need someone to quote. Fernhill's head roaster offers comment to food and business journalists, and is quoted by name.

YouTube and video

Mentions on YouTube had the strongest correlation with AI visibility in the Ahrefs study⁠(opens in a new tab) above, and YouTube was the most-cited website in Google's AI Overviews in Ahrefs' March 2026 data⁠(opens in a new tab). It's less prominent in ChatGPT's citations. As with the other studies, this shows a pattern, not that video causes AI mentions.

Useful videos for a business are product demonstrations, honest comparisons, how-to guides and interviews with experts. Fernhill films each brewing guide. Give each video a specific title, and divide it into chapters so viewers can jump to each part. Check that the automatic transcript YouTube creates is accurate. Then embed the video, which means putting its player on the matching page of your website. YouTube Help⁠(opens in a new tab) explains how to add chapters.

Appearing in other people's videos counts too. If you send a product to a reviewer or pay them, the review is an endorsement. It must say so clearly, under the FTC's rules⁠(opens in a new tab) in the US and the ASA's guide for influencers⁠(opens in a new tab) in the UK.

What to do next

1. List the websites that AI assistants cite for your five most important customer questions, and note which ones mention you.
2. Read the rules of every review site you use, then set up a routine that asks every customer for a review.
3. Find the three roundups or "best of" lists that appear most often, and contact their writers with accurate facts.
4. Choose one forum where your customers ask questions, and start taking part openly as your business.
5. Plan one piece of original research you could publish in the next three months.

Further reading

• Top brand visibility factors in ChatGPT, AI Mode and AI Overviews⁠(opens in a new tab), Ahrefs' study of 75,000 brands.
• Fake reviews: CMA guidance⁠(opens in a new tab), the UK rules on reviews.
• Consumer reviews and testimonials rule: questions and answers⁠(opens in a new tab), the FTC's explanation of the US rule.
• Wikipedia: Conflict of interest⁠(opens in a new tab), Wikipedia's rules for people writing about themselves or their employer.