You’ve spent three years collecting links, and your Domain Rating finally looks respectable. Then someone pastes a ChatGPT answer into the team chat, and your brand isn’t in it. A competitor with a thinner link profile is. That moment is why the backlinks AI search question has become so heated: are links still the currency, or did the rules change while we weren’t looking? I’ve gone through the major 2026 studies, from Ahrefs, Semrush, SE Ranking and Zyppy, to see where they agree and where they clash. The short version is more nuanced than either camp admits, and it changes what you should do next quarter. Let’s dig in together.

What the studies on backlinks AI search actually measure

Before any numbers, one thing trips people up: researchers measure different outcomes.

A mention is when an AI names or recommends your brand. A citation is when it links your page as a source, as Orbit Media’s citation tracking puts it. Some studies measure brand visibility in Google’s AI Overviews. Others count AI citations per domain in ChatGPT. Those aren’t the same thing, so the results can look contradictory when they aren’t.

Correlation numbers also need a quick translation. A figure near 0 means almost no relationship, and one near 1 means a strong one. Most results below land in the middle.

Ahrefs itself cautions that correlation does not establish causation, and that every factor it examined fell between moderate and very weak by conventional standards. You can read the original in Ahrefs’ brand visibility study. Keep that in mind for the rest of this article, because every backlinks AI search study has the same limit.

The case that links still matter

Referring domains keep showing up

When researchers test link metrics against citations, the number of different sites linking to you (referring domains) keeps surfacing.

SE Ranking analysed 129,000 domains and found referring domains were the strongest single predictor of ChatGPT citations, with average citations rising from 2.9 to 5.6 above 32,000 referring domains. A summary of the research walks through the method.

SALT.agency found moderate correlations between link metrics and citations: 0.39 for backlinks on ChatGPT, 0.42 for referring domains on Perplexity and 0.41 on Gemini, with a weak 0.25 for Domain Rating on Google AI Overviews. Their full write-up is worth a read.

A 2026 Australia and US study correlated domain-level citations from 760 AI responses against Ahrefs and Semrush data, and found referring domains correlated more strongly with citations than Google rankings or Domain Rating did. It’s a small sample, so treat it as a signal, not a verdict. The details are in the Earned Media study.

Why links might help indirectly

Links help pages get crawled, indexed and ranked, and AI tools lean on those indexes.

In Seer’s test, 87% of SearchGPT citations matched Bing’s top organic results. Google says it plainly too: to be eligible as a supporting link in AI Overviews or AI Mode, a page must be indexed and eligible to show in Search with a snippet, with no additional technical requirements (Google’s guidance).

Cyrus Shepard’s meta-analysis of 54 studies points the same way. It scored URL accessibility at 9.5 out of 10 and search rank at 9.4, while backlinks scored 4.8, described as present but indirect. His Zyppy write-up is the most consolidated view I’ve found.

So in the backlinks AI search picture, links often work through rankings and discoverability, not as a direct switch.

The case that links aren’t the main story

Brand mentions beat backlinks

This is the headline stat you’ve probably seen. In Ahrefs’ study of 75,000 brands, referring domains correlated 0.295 with AI Overview brand visibility and raw backlink counts 0.218, while branded web mentions hit 0.664 and YouTube mentions 0.737.

In plain English, being talked about tracks AI visibility more closely than being linked to. Brand mentions across independent sources look like the stronger signal.

That fits what I covered earlier in this series: Muck Rack’s May 2026 study of more than 25 million links found earned media made up 84% of AI citations, with paid or advertorial content at 0.3% (summary).

Ranking isn’t the ticket it used to be

Google’s AI Overviews now cite plenty of pages outside page one. Ahrefs found 38% of cited pages also appeared in the top 10, down from 76% in its July study (Search Engine Journal). BrightEdge’s tracking puts the overlap at about 17%, using a different method (BrightEdge).

Those figures disagree because they measure different things, but both say the same thing: a strong link profile and a top-10 ranking don’t guarantee a citation.

Thresholds and diminishing returns

Link strength seems to work more like a floor than a ladder. In SALT’s data, 18.5% of domains with a DR of 80 or higher fell into the lowest quarter of AI citations, while no domain with a DR of 40 or lower reached the top quarter.

Semrush’s 1,000-domain study adds a twist: nofollow links correlated with AI mentions about as strongly as followed ones (0.509 vs 0.504). See the Semrush study.

How to read backlinks AI search research without fooling yourself

Here’s how I’d sanity-check any claim in the backlinks AI search debate:

  • Is it a correlation or an experiment? Almost all of it is correlation.
  • Which outcome? Brand visibility, mentions and citations aren’t interchangeable.
  • Which platform? Studies disagree on ChatGPT, Perplexity, Gemini and AI Overviews, and only 11% of domains appear in both ChatGPT and Perplexity results.
  • How big is the sample? Some use hundreds of domains, others use tens of thousands.
  • Who paid for it? Many publishers sell link building or AI-visibility tools.
  • How old is it? Google’s Gemini 3 upgrade in January 2026 changed AI Overview behaviour, so 2025 results may be stale.

Step by step: a practical backlinks AI search plan

1. Make sure the bots can reach you

Check your robots.txt. OpenAI documents that OAI-SearchBot is the crawler that surfaces sites in ChatGPT search, and sites opted out of it won’t be shown in ChatGPT search answers. Crawlability is the cheapest win in the whole list.

2. Keep earning relevant links

Don’t abandon link building. Aim for relevant referring domains from real sites, not volume. Links still drive rankings, and rankings still feed citations.

3. Get mentioned where AI reads

Earn coverage, reviews and community mentions. SE Ranking found domains present on several review platforms averaged 4.6–6.3 citations versus 1.8 for those on none. Digital PR, expert quotes and honest participation in communities all help.

4. Publish something worth citing

Original data is among the most consistent signals in RankSense’s review of the studies. Write clear, self-contained passages that answer questions directly.

5. Track it weekly

Pick ten buyer-style questions, run them in ChatGPT, Perplexity and Gemini, and log whether you’re mentioned or cited. Do it for a month before drawing conclusions.

Mistakes in a backlinks AI search strategy

  • Stopping link building because of one headline number. The 0.218 figure is about one outcome in one study.
  • Buying links for AI visibility. Paid placements barely register in the data.
  • Chasing DR. Metrics are a filter, not a goal.
  • Treating llms.txt as a shortcut. Zyppy’s analysis scored it 2.0 out of 10, the lowest of all factors.
  • Assuming a citation means credit. One Semrush analysis found 62% of AI citations never name the brand behind the page.
  • Optimising for one platform. Each engine behaves differently.

Final thoughts

So, do backlinks influence AI search citations? The honest answer is “yes, but mostly in the background.” Links help you get found, ranked and trusted, while brand mentions, relevance and clear content seem to do more of the direct work.

If you only change one thing this month, run the weekly prompt test. A few weeks of real data on your own brand will tell you more than any single backlinks AI search study, including this one.

Sources: