It's the question everyone building online is asking right now: if people are getting answers from ChatGPT, Perplexity, and Google's AI Overviews instead of clicking blue links, do backlinks even matter anymore? The short answer is yes—but the reasons are worth understanding, because they change how you should think about earning them.

Why the Question Comes Up

The logic behind "backlinks are dead" goes like this: backlinks were a proxy for what pages humans should see in search results. If AI assistants generate answers directly, there are no rankings to climb and no clicks to win, so links must be irrelevant.

That reasoning misses how these systems actually work. AI answers don't come from nowhere. They're built on top of the same web that backlinks have always described—and the signals that made a page authoritative for search still make it authoritative for a language model deciding what to trust.

How AI Systems Actually Use the Web

There are two mechanisms at play, and both lean on authority.

Training data. Large language models learn from a massive crawl of the web. Pages that are widely linked and frequently cited appear more often, get referenced by more sources, and are represented more strongly in what the model "knows." A brand mentioned across hundreds of relevant sites becomes part of the model's understanding of a topic. A brand mentioned nowhere effectively doesn't exist to it.

Retrieval at query time. Tools like Perplexity and Google's AI-powered search don't rely on training memory alone. When you ask a question, they retrieve live pages and summarize them. That retrieval step ranks candidate sources—and rankers lean heavily on the same authority and citation signals that classic search uses. Well-linked, frequently-cited pages surface first, so they're the ones the model reads and paraphrases.

In both cases, backlinks aren't a legacy artifact. They're one of the clearest signals a machine has for deciding which sources deserve weight.

Links Aren't the Whole Story Anymore

Here's the nuance that matters: for AI visibility, unlinked brand mentions count too.

A traditional backlink passes a signal through an HTML link. But an LLM reading the web doesn't need a clickable link to associate your brand with a topic. If a respected industry article says "tools like Referivo handle outreach research automatically," that sentence teaches the model something—even with no link attached.

This is a real shift. In classic SEO, an unlinked mention was worth little because there was no link equity to pass. For AI systems learning from text, the mention itself is the payload. The model reads the words, forms associations, and can surface your brand later.

That doesn't make links worthless. Links still drive retrieval ranking, still send referral traffic, and still help human readers and crawlers find you. But it does mean you should stop thinking purely in terms of "did I get a dofollow link" and start thinking in terms of "am I being mentioned in the right places, in the right context."

What Changes Versus Classic SEO

If you've done link building before, most of the fundamentals carry over. A few emphases shift.

Relevance and context matter more. An LLM pays attention to the words around your brand. Being described accurately, next to the right topics, shapes how models associate you. A link buried in an irrelevant directory does little; a sentence that explains what you do, on a page about your category, does a lot.

Citation frequency compounds. Being mentioned once is noise. Being mentioned repeatedly across credible, independent sources builds a pattern the model recognizes. Consistency of association is what turns "a company" into "the company people mention for this."

The destination isn't just a ranking. With classic SEO, the payoff was a position on a results page. With AI search, the payoff is becoming part of the answer itself—and part of the model's baseline knowledge. That's a durable asset that doesn't reset with every algorithm update.

Vanity metrics matter even less. Chasing raw link counts or a single authority score was always shaky. For AI visibility it's worse, because a pile of low-quality links teaches a model nothing useful. Quality and relevance of the mention dominate.

This is the core idea behind generative engine optimization: optimizing not for a ranking slot, but for how AI systems learn about and cite your brand. If you want the full comparison of what carries over and what doesn't, we broke it down in GEO vs. SEO.

A Practical Approach for Founders

None of this requires a new playbook built from scratch. It requires pointing the same energy at a slightly wider target: earned mentions, not just links.

Pitch people, not URLs. The old game was submitting links to pages. The durable game is getting a real person—a journalist, a podcast host, a newsletter writer, an industry blogger—to mention what you do because it's genuinely relevant to their audience. Those mentions are what both search rankers and language models learn from.

Prioritize relevance over raw authority. A mention on a focused industry site your buyers actually read teaches models more useful associations than a link from a high-authority site with no topical connection. Chase context, not just domain scores.

Aim for repeated, consistent coverage. One mention won't move the needle. A steady pattern of being referenced across credible sources in your category will. Think of it as building a reputation the models can observe, not a one-time campaign.

Get the context right. Since the words around your mention do the teaching, care about how you're described. When you pitch, make it easy for the writer to describe you accurately and specifically.

Track the right outcomes. Rankings and referral traffic still matter, but also start checking whether AI assistants mention you. Ask ChatGPT, Perplexity, and Gemini questions your customers would ask, and see whether your brand shows up. That's your new visibility signal.

Where Referivo Fits

Doing all of this by hand is slow. You have to find the places you should be mentioned, figure out who writes there, dig up their contact details, and craft a pitch that's actually worth a reply—for every single opportunity.

That's the work Referivo automates. It finds where you should get mentioned, surfaces the authors and their contacts, and drafts personalized outreach so you're pitching people, not blasting URLs. The earned mentions and citations that come out of it are exactly the signals that help you rank higher on Google and show up when someone asks an AI assistant for a recommendation in your space.

The Bottom Line

Backlinks still matter for AI search—just not in isolation, and not as a number to inflate. What matters is being mentioned, accurately and repeatedly, in the credible places your audience and the models both pay attention to. Links remain part of that. Unlinked mentions now count alongside them. Both feed the reputation AI assistants learn brands from.

The founders who win the next few years won't be the ones with the biggest backlink count. They'll be the ones the models have learned to trust because the right people kept mentioning them. Get mentioned, and you grow faster on both fronts at once.