AI search and links: which sources it trusts and cites

AI search does not trust a link according to a fixed, known rule. All that is publicly documented is this: chat assistants name their sources, and Google's AI Overviews and AI Mode link to web pages. The providers do not disclose how they select sources in detail. Even so, a pattern can be observed, and you can draw sensible decisions from it.

In practice, there is often a sequence. A page must first be found. Then it must look reliable enough. Finally, it must give a clear statement that fits the question. Authority opens the door. Mentions, topical fit, variety and freshness influence which brand ends up in the answer. This article explains that mechanism and what it means for choosing your media.

How AI search preselects sources

Many AI answers are produced in two steps. First, the system retrieves relevant documents. This is known as retrieval. Then a language model writes an answer from those documents and, depending on the product, shows the sources.

This has a consequence for you: the model can only cite what ends up in the retrieval results. Pages that retrieval never finds do not appear in the answer. The preselection therefore happens before the actual writing.

In practice, cited pages are often also highly visible in classic search. You should not read this as a law. It is a plausible observation: a page that already ranks well for a question is also more likely to be found as a candidate. For an introduction to the terms, see our article What is GAIO and why mentions are the lever.

Authority as a trust threshold

Classic SEO often treats authority like a league table: more signals, better position. With AI answers, authority works more like a threshold. A domain needs enough editorial signals to be considered a candidate at all. These include mentions and links from credible media, a clean profile and recognisable topical expertise.

Clearing the threshold does not mean you are cited. But if you do not clear it, you will struggle to appear at all. Editorial links therefore remain an important building block. The article Editorial links: building authority through editorial mentions explains why they build authority.

Mentions with and without a link

Within the pool of qualified sources, context plays a bigger role. Language models read text. They see which brand is named in which context, even when there is no clickable link. An editorial mention with a clear statement about your company can therefore be valuable for AI answers, even if the link is set to nofollow.

This is not proven in every detail for individual systems. As an industry observation, however, it is consistent: where a brand is named repeatedly in fitting contexts, it creates a brand signal that a plain link count does not capture.

This is no argument against links. The two signals complement each other:

• Links from editorial sources support rankings and trust in the domain.
• Mentions create the brand context that a model can use when writing an answer.

The reference Link attributes in SEO describes what the rel attributes mean technically. Paid placements must be labelled as advertising. In its spam policies, Google requires paid links to be marked with sponsored or nofollow. A placement that is honestly labelled can still build visibility and brand context.

Variety and freshness

When several independent sources say the same thing about a brand, it carries more weight than a single large placement. In practice, a broad picture across different reputable media often helps more than many mentions on the same domain.

Three points matter here:

• Topical fit: An outdoor retailer benefits more from a mention in an outdoor magazine than in a general directory.
• Different domains: Several independent media provide a more varied signal than repeated mentions in one place.
• Freshness: A maintained, up-to-date article is more useful as a source than a text that has sat unchanged for years. This also applies to the page being linked to.

The target topic itself is a special case. If your offer is seasonal, plan publications to appear before demand peaks. Digital PR and brand mentions describes patterns from successful campaigns.

What this means for your media choice

The mechanism suggests a clear order. First build authority, then secure topical fit and variety, and finally make sure your own site offers current, clear statements. For choosing media, this can be turned into a checklist:

Note that metrics are aids, not guarantees. A domain with good values can still be the wrong one for your topic. Always check the outlet's editorial presence as well.

The publisher decides the link type according to their editorial guidelines. Clarify in the briefing what you can expect. The article DoFollow vs. nofollow: what advertisers need to know helps here. If you already have a text or a press release, the content assistant suggests suitable channels and explains why.

mention.media does not promise specific ranking results or citations in AI answers. You can, however, improve the conditions: fitting media, honestly labelled formats and clear messages.

Frequently asked questions

Does an AI only cite pages that already rank well?

No, this is not a fixed rule. In practice, though, cited sources are often also highly visible in classic search. Plan for authority as a foundation.

Does a mention without a link help AI visibility?

It can. Language models read text, and that includes brand mentions in context. There is no guarantee. A mention in a fitting, credible outlet is still a sensible building block.

Do I have to label paid mentions?

Yes. Paid placements must be labelled as advertising. For links, Google's spam policies require rel sponsored or nofollow. This protects you and the outlet.

How do I find suitable media?

In the catalogue, use the filters for media type, format, niche, maximum price and SEO metrics. Without signing up, you see a demo extract with anonymised publishers. Your first booking in four steps shows how to get from selection to booking.