Unlinked Brand Mentions Now Do What Links Used To

For twenty years the answer to “how do I build authority” was links. Get enough good ones and rankings follow.

That still works. But something has been added alongside it, and it behaves differently enough that treating it as link building with extra steps will produce the wrong strategy.

AI systems answering questions about your market do not consult a link graph in the way a traditional ranking system does. They draw on what has been written about you across many sources, and consistency across those sources is doing work that brand mentions never used to do on their own.

The Mechanism, Plainly

When a language model is asked to recommend a provider in your category, it is not running a query and ranking results by authority score. It is drawing on patterns in what it has read, and on whatever sources it retrieves at the moment of answering.

Both of those reward the same thing: being described, by other people, in consistent terms, in enough places.

A link is a vote that carries authority. A mention is a description that carries meaning. For a system trying to establish what your company is and whether it should be named, the second one is often more directly useful, and it does not need to be a link at all.

This is why companies with modest link profiles sometimes appear in AI answers ahead of much larger competitors. They have been discussed specifically, in relevant places, in consistent language.

Where Those Mentions Live

Discovery has fragmented across platforms in a way that traditional SEO reporting mostly does not capture. The places people research now (and therefore the places that get read and cited) include Reddit, YouTube, LinkedIn and TikTok alongside conventional publications.

Reddit deserves particular attention. It is heavily used as a research surface by people who distrust marketing content, it is heavily represented in search results, and threads asking for recommendations in a category are exactly the kind of source an AI system retrieves when asked the same question.

None of which means you should go and post about yourself on Reddit. That is transparent, generally against community rules, and reliably backfires. It means that whether your product gets named in those conversations by actual users is now a marketing outcome worth caring about, and it is downstream of the product being good and known, not of anything an SEO can execute directly.

That is an uncomfortable finding for the SEO industry and I think it is a true one.

What You Can Actually Influence

Three things, in descending order of how much control you have.

The first is the most underrated and the cheapest. If your own site calls you an SEO agency, your LinkedIn says growth marketing consultancy, and a directory lists you under web design, you have given every system reading you three incompatible descriptions and no reason to prefer any of them. Inconsistency is not neutral, it is actively corrosive.

Fixing it is an afternoon of work across your profiles and structured data. The entity side of that (describing yourself once, properly, in markup) I have covered in schema markup for AI search.

The Third One Is the Real Constraint

Being nameable is the part that cannot be outsourced to an SEO, and it is usually the binding constraint.

Ask what an AI system would say if someone asked why they should choose you. If the honest answer is “they do SEO, like the other forty results”, there is nothing to retrieve and nothing to state. No amount of digital PR fixes a positioning problem.

The companies that get named consistently tend to have one specific, verifiable thing attached to them: a niche, a method, a dataset, a founder known for something. That is a business decision rather than a marketing one, and it is the reason this cannot be reduced to a service line.

On the Statistics Going Around

You will see confident figures attached to this topic: percentages of searches ending without a click, proportions of brand queries showing AI Overviews, shares of marketers tracking AI visibility. Most of them trace back to vendor studies and industry surveys rather than primary research, and they get repeated between blogs until they acquire an authority the original methodology does not support.

The directional claims are almost certainly right. Zero-click behaviour is up substantially. AI surfaces appear on a large share of commercial queries. Very few companies measure any of it. I would rely on the direction and treat any specific decimal place with suspicion, including in articles arguing for services I sell.

Your own Search Console generative AI report, now that it exists, is worth more than every industry statistic combined, because it describes you.

What I Would Do

1. Audit your descriptions. Every profile, directory and bio. Make the category and the claim identical across all of them.
2. Find where you are already mentioned. Search your brand name and see what exists, what is accurate, and what is years out of date.
3. Fix the wrong ones. Outdated descriptions on third-party sites are actively teaching systems something untrue about you.
4. Earn a small number of good mentions in places that genuinely get read in your industry. A handful of relevant ones beats volume comfortably.
5. Decide what you want to be known for before doing any of the above, because the rest amplifies whatever that is.

Step five is the one that takes months and the one nobody wants to do. It is also the only one that changes the outcome rather than the inputs.

Written by Shahzaib Ul Hassan, senior AI SEO consultant and founder of ShazzSEO. Ranking sites since 2009. 500+ websites optimized, 3,000+ students trained.