Backlinks vs. Brand Mentions: What Actually Earns AI Citations

For AI citations, the brand mention usually does more of the work than the link. Answer engines read the text of the pages they retrieve, so a clear, specific mention of your brand on a page that gets retrieved can put you in the answer whether the link is dofollow, nofollow, sponsored or missing entirely.

The link still earns its place, just elsewhere: classic search rankings, crawler discovery, referral traffic, and making it unambiguous which company a name refers to. Those effects feed back into AI visibility, which is why the useful answer is rarely "mentions instead of links". It's knowing which one a placement is buying you, and why.

This guide separates the two, explains why rel attributes don't change what a model reads, and gives a simple way to decide what to prioritize for each placement. For the bigger picture of how placements turn into citations, start with how backlinks help your brand get cited.

Five kinds of reference, defined

Marketers often lump these together. They behave differently.

• Dofollow link: an ordinary hyperlink with no qualifying rel value. Search engines can follow it and count it as an endorsement for ranking.
• Nofollow link: marked rel="nofollow". Google suggests it when you don't want your site associated with, or crawlers following, the linked page.
• Sponsored link: marked rel="sponsored", which Google asks sites to use for ads and paid placements. rel="ugc" is the equivalent for user-generated content such as comments and forum posts.
• Linked mention: your brand named in the text, with a link of any type on or near the name.
• Unlinked mention: your brand named in the text with no link at all.

Google's documentation says links carrying these rel values "will generally not be followed" (Google Search Central). That's a statement about crawling and ranking in Google Search. It says nothing about whether an AI system reading the page notices your brand name, because that is a different process.

Why answer engines read text, not rel attributes

When an engine retrieves a page to answer a question, the part that reaches the model is the content: headings, paragraphs, lists, maybe a table. The model reads a passage like "Acme is an invoicing tool for UK freelancers that files VAT returns" and can repeat it, citing the page. The rel attribute on the hyperlink around "Acme" isn't part of what it weighs.

That's an inference about how these systems are built, but a well-grounded one. A rel value is an instruction for crawlers and ranking systems, while the generation step works on language. Nothing in the public documentation from OpenAI, Perplexity, Google, Microsoft or Anthropic suggests that a sponsored or nofollow link causes the surrounding text to be ignored.

Link signals can matter upstream, in the search index that decides which pages get retrieved. Engines grounded in Google or Bing inherit link-based ranking, so the links pointing to the publication's page affect whether it's retrieved. The link going out from that page to you is a different matter: its rel value doesn't change whether the page itself gets retrieved or what it says.

Co-occurrence: what surrounds the mention matters most

If the text is what counts, the words around your brand name are doing the real work. Many of the prompts that matter commercially are category questions ("best invoicing tools for freelancers") rather than questions about a brand the user already knows, so a mention helps most when it sits next to the category terms and attributes buyers use in those prompts.

Compare two mentions of the same hypothetical brand:

• Weak: "We've been really impressed with Acme lately."
• Strong: "Acme is an invoicing tool for UK freelancers that files VAT returns and chases late payments automatically."

The second puts the brand in one sentence with its category (invoicing tool), its audience (UK freelancers) and two concrete capabilities. A retrieval system matching a freelancer-invoicing query has something to match, and a model writing the answer has a sentence it can lift. Repeated across several publications, that pairing is also how a brand becomes associated with a category over time.

A strong mention usually contains:

• The brand name, written the same way every time
• The category, in the plain words a buyer would type
• Who it's for
• One or two specific, checkable facts
• Context that frames it as a recommendation or comparison, not a slogan

How to build whole articles around passages like this is covered in writing placement articles that get quoted by AI.

When the link itself still matters

None of this makes links pointless. A link does several jobs that a mention can't.

Classic rankings for your own pages

A followed link from a relevant, authoritative publication passes ranking signals to the page it points to. That helps your product, comparison and pricing pages rank, and ranking pages are what AI engines retrieve when someone searches for you or your category. Nofollow and sponsored links contribute little here, since Google says it generally won't follow them.

Discovery and crawling

Links are one of the main ways crawlers find pages. A new product page or a fresh piece of research tends to be found and indexed faster when established sites link to it, and only an indexed page can be retrieved.

Referral traffic and measurement

When an AI answer cites a publication, some readers click through to it, and some of those click on to you. That path only exists if there's a link. It also gives you something concrete to track, which matters because AI visibility is otherwise hard to measure; measuring AI citations from backlinks explains how to use it.

Disambiguation

If your brand shares a name with something else, a link to your site removes the doubt for any reader, human or machine, working out which company the text means. It's plausible that this helps engines attach a mention to the right entity, though none of them document it.

Disclosure isn't a weakness

Google's spam policies list buying or selling links for ranking purposes as link spam, and state that paid links are not a violation when they're qualified with rel="sponsored" or rel="nofollow". Publishers handle this differently: some label sponsored content and qualify its links, others don't.

For AI citations, a properly labeled sponsored placement gives up very little, because the text is what gets read. It also keeps the publisher in good standing with search engines, which protects the authority you're borrowing when you place an article there. InTheAnswer's AEO score method factors link type and sponsored labeling into its link-hygiene signal, alongside whether a site's traffic is consistent with its authority. The ways link buying commonly backfires are covered in backlink mistakes that hurt AI visibility.

Deciding what to prioritize

For any single placement, the useful question isn't "link or mention?" but "what does this placement need to do?" A simple sequence:

1. If you need to appear in category answers, prioritize strong mentions on publications that already get cited for your category, and accept nofollow or sponsored links if that's what those publications offer.
2. If you need your own pages to rank, followed links from relevant publications are what move classic rankings. Links earned editorially raise no policy question; for paid placements, keep Google's policy above in mind.
3. If your brand name is ambiguous, insist on a link from the mention to your homepage, whatever its rel value.
4. If a site offers a dofollow link but has no topical fit, treat it as a classic-SEO purchase at best; it's unlikely to be retrieved for your category prompts.
5. If you'll need to report results, make sure every placement includes a link you can track.

Most brands end up with a mix. The catalog shows the link type on every listing, including 2,569 dofollow placements, so you can plan the mix before you buy rather than discover it after publication.