Do Backlinks Still Matter for AI Search?

Backlinks still matter, but less than your link-building budget assumes. Across 75,000 brands, Ahrefs measured backlinks at a 0.218 correlation with AI visibility and brand web mentions at 0.664. Links help you rank, and rankings still feed some AI answers. They are no longer the main thing AI reads when it decides who to cite.

If you run an SEO program, this is the question your team is quietly arguing about. Half the budget goes to link building. Buyers now open ChatGPT, Perplexity, and Google AI Overviews before they ever see a blue link. So the spend deserves a hard look: does the link still buy what it used to?

The honest answer has two halves. Backlinks have not stopped working. They have stopped being the lever. This post separates the two so you can move money to where it earns now.

The 30-second answer

Backlinks still matter for AI search, but weakly and indirectly. Ahrefs found they correlate with AI Overview visibility at just 0.218, the lowest of every brand signal it measured. Brand web mentions correlate at 0.664. Links help you rank, which still feeds Google AI Overviews. They do almost nothing for ChatGPT and Perplexity, which read mentions, not link graphs. Keep link building for rankings. Fund brand mentions for citations.

Do backlinks correlate with AI citations at all?

Yes, but they sit at the bottom of every ranked signal. The largest public study on this comes from Ahrefs, which analyzed 75,000 brands in May 2025 to find what predicts inclusion in Google AI Overviews. Backlinks landed last.

Backlinks rank dead last among brand-visibility signals

Ahrefs measured five factors against AI Overview visibility. Brand web mentions came first at 0.664. Brand anchors followed at 0.527, then brand search volume at 0.392. Referring domains reached 0.295. Backlinks finished last at 0.218.

Read that order again. The three strongest predictors are all off-site reputation signals. The two link metrics, referring domains and backlinks, are the weakest of the set. A link is a vote one site casts for you. A mention is the web agreeing you exist.

Referring domains only move the needle at enormous scale

There is one place links clearly help, and it is a scale most companies will never reach. SE Ranking analyzed 129,000 domains across 216,524 pages to rank the factors behind ChatGPT citations. Referring domains came out as the single strongest predictor in that dataset, but with a sharp threshold.

Sites with up to 2,500 referring domains averaged 1.6 to 1.8 citations. The jump came at 32,000 referring domains, where citations roughly doubled. Sites above 350,000 referring domains averaged 8.4 citations. For context, almost no B2B SaaS site has 32,000 referring domains. That tier is Wikipedia, Reddit, and major news publishers, not your product blog.

Brand mentions predict AI visibility about three times better than links

The gap between mentions and links is not subtle. At 0.664 versus 0.218, brand mentions predict AI Overview visibility close to three times more strongly than backlinks. Ahrefs also found that brands in the top quartile for web mentions averaged 169 AI Overview mentions, more than ten times the next quartile down.

The unit of value moved. SEO counted links into your site. AI search counts mentions of your name across the web, with or without a link attached.

Why backlinks lost their grip on AI answers

The shift is mechanical, not philosophical. AI answer engines do not traverse a link graph the way Google's classic crawler does. They retrieve passages and repeat claims. A link is invisible to that process unless it also changes what the web says about you.

AI retrieval reads passages, not link graphs

When you ask ChatGPT or Perplexity a question, the model does not score pages by how many sites link to them. It pulls the cleanest passage that answers the prompt and attributes it. The link economy that powered PageRank for two decades is not the input. The text is. That is the core idea behind how AI citations actually work, and it is why a thin page with a sharp answer can beat a heavily linked page that buries its point.

AI does not follow your links. It repeats what the web already says about you.

ChatGPT and Perplexity pull from Reddit, Wikipedia, and editorial, not your backlink profile

The sources these engines lean on are public and consistent. Ahrefs research on ChatGPT citations found Reddit and Wikipedia alone account for close to half the citation pool, before you count YouTube, category review sites, and trade press. Ahrefs' follow-up study across ChatGPT, AI Mode, and AI Overviews found YouTube mentions were among the strongest signals of AI visibility of all.

None of those surfaces care about your domain rating. They care whether your name appears, in context, where the model is already reading. We mapped that source list in detail in our breakdown of the top domains AI search cites.

The one place links still count: Google AI Overviews

Google AI Overviews are the exception, because they sit on top of Google's own index. Cited pages there skew heavily toward pages that already rank, and backlinks still help pages rank. So a link can reach an AI Overview citation through the back door of ranking, even though it does almost nothing on ChatGPT or Perplexity.

This is the split most teams miss. Backlinks build the rankings AI reads. They do not build the reputation AI repeats.

A link builder asks:

• •How many referring domains does this page have?
• •What is the domain rating of the linking site?
• •How do we earn a dofollow link this quarter?

An AI search strategist asks:

• •How often is this brand named across the open web?
• •Do Reddit, Wikipedia, YouTube, and trade press mention it?
• •Can a model lift a clean, factual claim about it?

Where backlinks still earn their budget

This is not an argument to fire your link team. Links still do real work. The point is to spend on them for the right reason and stop expecting them to drive AI citations directly.

Backlinks still drive the Google rankings that feed AI Overviews

Google AI Overviews and AI Mode reach more than two billion users a month combined. They draw from Google's ranked index. If links lift your rankings, they lift your odds of surfacing in those answers. For any brand whose buyers use Google's AI surfaces, ranking work remains a legitimate AI visibility investment, just an indirect one. We covered the rank-to-citation relationship in why your Google ranking no longer guarantees an AI citation.

High-authority links get you into the source pool, not the citation

A link from a publication AI already cites can do something a generic link cannot: it puts a sentence about you on a page the model reads. The value is not the link attribute. It is the mention that rides along with it. A single bylined feature on a trade publication that AI cites beats fifty directory links that no model will ever retrieve.

The win is the placement, not the hyperlink. The same earned-media work that lands the mention happens to land the link.

Link building and brand mentions are not the same budget line

Most teams file both under one heading and let link building eat the whole thing. Separate them. Link building chases the hyperlink. Mention building chases the named reference, linked or not. Up to 85% of brand references in AI answers carry no clickable link at all, which means a mention-only strategy still wins citations a link-only strategy would miss. The fastest AI visibility gains we see come from the work in our off-page citation playbook, where a brand with zero domain authority reached top-three in ChatGPT in 14 days on mentions alone.

How to reallocate a link-building budget for AI search

Here is the prescription half. If your link budget is built for a search world that ran on hyperlinks, retune it for a world that runs on mentions. Five moves, in order.

1. •Audit where your name appears today, not where your links point. Run 20 buyer-intent prompts in your category across ChatGPT, Perplexity, and AI Overviews. Log every cited domain. That list is your real target set, and it is rarely the set your outreach team has been pitching.
2. •Hold link building only for ranking-driven surfaces. Keep earning links to the pages whose Google rank you need, since rank still feeds AI Overviews. Stop measuring those links by referring-domain count and start measuring them by whether the host page is one AI cites.
3. •Move the freed budget into earned brand mentions. Fund placements on the sites the audit surfaced: Reddit threads, Wikipedia and Wikidata records, YouTube, category review sites, and the two or three trade publications AI repeats for your category.
4. •Build a clean entity record so mentions connect to one brand. Organization schema, a Wikidata entry, and consistent naming across the web let models tie scattered mentions to a single entity instead of treating them as unrelated text. This is the foundation under everything in generative engine optimization.
5. •Measure citation share, not link velocity. Track how often AI names your brand on a fixed prompt set, weekly, by platform. Links are now a leading input to one surface. Citation share is the outcome that pays.

If your brand is missing from AI answers entirely, start with the broader diagnostic in why your brand is not showing in ChatGPT, then come back to the budget split.

FAQ

Do backlinks help with ChatGPT and Perplexity?

Barely. Both engines retrieve passages and repeat claims rather than scoring pages by inbound links. Ahrefs measured backlinks at a 0.218 correlation with AI visibility, the weakest of every signal tested. What moves these engines is brand mentions on the sources they read, mainly Reddit, Wikipedia, YouTube, and editorial publishers. Spend on the mention, not the link.

Are backlinks still important for SEO in 2026?

Yes, for ranking. Backlinks remain a real ranking factor in Google's classic index, and rankings still feed Google AI Overviews. The change is that links no longer drive AI citations directly. Keep link building for the pages whose rank you need, and stop expecting it to win ChatGPT or Perplexity citations on its own.

What predicts AI citations better than backlinks?

Brand mentions. Ahrefs found brand web mentions correlate with AI Overview visibility at 0.664, brand anchors at 0.527, and brand search volume at 0.392, all far above the 0.218 for backlinks. The pattern holds across studies: how widely the open web talks about you predicts AI visibility better than how many sites link to you. See our analysis of the strongest predictor of AI citations.

How many backlinks do you need to get cited by AI?

Far more than most sites can realistically build. SE Ranking's analysis of 129,000 domains found citation odds only jumped meaningfully at 32,000 referring domains and kept climbing past 350,000. That tier belongs to Wikipedia, Reddit, and major publishers. For a typical B2B site, chasing that volume is the wrong goal. Earning mentions on those high-authority domains is the better one.

Should I stop link building and only build brand mentions?

No. Split the budget. Keep enough link building to hold the Google rankings that feed AI Overviews, then move the rest into earned mentions on the sites AI cites. A link-only strategy misses the roughly 85% of AI brand references that carry no clickable link. A mention-first strategy captures both.

The bottom line

Backlinks did not die. They got demoted. They still build the rankings that feed Google's AI surfaces, and a link on a page AI reads carries a useful mention with it. What they no longer do is drive citations on the engines that read text instead of link graphs.

The teams winning AI visibility in 2026 stopped measuring referring domains as the goal and started measuring how often the open web names them. Move part of the link budget there, keep the rest pointed at rank, and measure citation share weekly. The full operating cadence lives in our AEO 101 playbook, refreshed daily from the research we publish.