Do backlinks still matter for AI Overviews?

Yes, backlinks still matter for AI Overviews, but as a supporting signal rather than the lever. The most reliable 2025 data shows backlinks correlate weakly with AI Overview visibility (Spearman r about 0.22), while branded mentions show a correlation about three times higher (r about 0.66). Links help you rank in the underlying index that AI Overviews draw from. They no longer guarantee a citation.

That nuance is the whole game right now. For a decade the playbook was simple: earn authoritative links, climb the rankings, capture the click. AI Overviews broke the last step of that chain. Google now synthesizes an answer at the top of the page, cites a handful of sources, and the pages it cites are not reliably the ones sitting in the classic top 10. So the question practitioners keep asking is not academic: if a backlink no longer buys a citation, is link building still worth the budget in 2026? Below is what the studies actually say, what changed, and how we'd allocate a link-building budget today.

What the data actually says about backlinks and AI Overviews

The most-cited dataset on this question is Ahrefs' analysis of 75,000 brands, published 26 May 2025, which measured the Spearman correlation between various signals and brand visibility inside AI Overviews. The headline: branded web mentions correlated far more strongly than any link metric. You can read the full breakdown in Ahrefs' AI Overview brand correlation study.

Read that table the way Ahrefs intends it. The raw number of backlinks sits near the bottom at 0.218. But Domain Rating (0.326) and referring domains (0.295) are link-derived metrics and they outrank traffic and most paid signals. So links are not irrelevant. They are simply outclassed by mentions. As Matt Diggity summarised the study, web mentions tell an AI model what to trust, while a backlink tells it where to navigate. Other studies do not all agree. Semrush's separate analysis of 1,000 domains, published in October 2025, found Authority Score, its link-quality metric, correlated with the number of AI mentions at 0.65 (Pearson), but only about 0.23 with a brand's share of voice in AI answers (Semrush).

Why links still matter, just indirectly

Google has been consistent that AI Overviews and AI Mode are not a separate ranking engine bolted onto Search. In its official guidance, the Search Relations team (John Mueller) stated that Google's long-standing advice still applies: create unique, satisfying content for people, and the same systems that rank pages organically feed the AI experiences. That guidance is documented in Google's official advice on optimizing for AI Overviews and AI Mode via Search Engine Journal.

The mechanical consequence is simple. To be eligible for an AI Overview citation, your page generally has to be indexed and ranking somewhere credible for the relevant sub-queries. Links remain one of the strongest ways to get a page crawled, indexed, and ranked in the first place. No links, weak rankings, less chance of being in the candidate pool the AI selects from. This is why the link/AI debate is not binary. Links are upstream of citations even when they don't directly cause them.

This is also why we steer clients toward link strategies that double as brand-building. A digital PR campaign that lands coverage in a trade publication earns a link and a branded mention in the same hit, which is exactly the combination the Ahrefs data rewards. Compare that with a thin guest post whose only payload is an exact-match anchor: it nudges DR slightly and produces no mention at all.

The overlap between rankings and AI citations is shrinking

Here is the uncomfortable trend for anyone who equates "rank #1" with "get cited." The share of AI Overview citations coming from top-10 organic pages has fallen steeply. In Ahrefs' July 2025 study, about 76% of the pages cited in AI Overviews also ranked in the top 10; in its March 2026 update, run on 863,000 keyword results with better citation parsing, the share was 37.9%, and the rest split almost evenly between positions 11 to 100 and pages outside the top 100 (Ahrefs). BrightEdge, tracking nine industries from May 2024 to September 2025, found 54% of citations overlapping with pages that rank somewhere in the results, but only 16.7% coming from the top 10 (BrightEdge). The numbers vary by methodology, but the direction is clear: AI Overviews cite many pages from beyond the top of page one.

For link builders, the strategic implication is to stop pointing every link at one money page. Spread authority across a topical cluster so multiple URLs are competitive for the sub-queries the fan-out generates. That is where disciplined internal linking strategy earns its keep: it distributes the equity your external links bring across the whole cluster instead of stranding it on one page. If you want the broader strategic frame, our guide to generative engine optimization (GEO) covers how to structure content for fan-out retrieval end to end.

How to split a 2026 budget between links and mentions

You do not have to choose. The smartest allocation treats links and mentions as one motion. The table below is how we'd frame the trade-offs for a mid-market brand spending on off-page work this year.

Notice that the tactics with the highest combined value are the ones that produce a link and a named mention together. That is the practical takeaway from the correlation data. If you're a SaaS brand specifically, the mechanics of doing this cleanly are in our guide to white-hat link building for SaaS, and if HARO's decline left a gap in your expert-quote pipeline, see HARO alternatives for 2026.

How this fits AEO, GEO, and classic SEO

Links are now one input into a broader system of entity authority. The other inputs, structured data, brand mentions, semantic completeness, and answer-shaped content, sit under the umbrella of answer engine optimization. The relationship between optimising for answers and optimising for rankings is not a replacement, it's a layering. We unpack exactly where they diverge in AEO vs SEO, and the underlying mechanics of how engines pick sources are covered in our GEO guide. For definitions, our glossary entries on backlink and domain rating keep the terminology straight.

There is also a commercial upside that makes the work pay for itself even amid the zero-click panic. Being cited inside an AI Overview is not pure charity: studies tracked by industry analysts found cited pages earn meaningfully more organic clicks than uncited competitors for the same query. The citation itself becomes a visibility asset. For the wider numbers on how AI search is reshaping click-through and referral patterns, see our regularly updated link building and AI search statistics, and benchmark what authoritative placements actually cost in the ANGLE link pricing index.

What to actually do this quarter

1. Audit which of your pages already appear in AI Overviews and which sub-queries trigger them, so you know your current citation footprint.
2. Reallocate part of your raw guest-post budget toward digital PR and expert-quote tactics that generate named mentions alongside links.
3. Map your money topics into clusters and earn or build internal links so several URLs compete for the fan-out sub-queries, not just one.
4. Rewrite key pages to include self-contained, statistic-backed sentences an AI can quote verbatim.
5. Track DR, referring domains, and unlinked mentions together, treating them as one entity-authority dashboard rather than separate KPIs.

The honest summary: backlinks have moved from being the headline act to being part of the supporting cast. They still get you into the room. Whether you get quoted on stage now depends on how recognisable and citable your brand is once you're there. Build links that build your name, and you satisfy both the old algorithm and the new one.