The short answer, with the numbers
Brand signals beat backlinks. By roughly three to one.
Ahrefs studied 75,000 brands and measured the correlation between off-site signals and AI Overview citation visibility. Branded web mentions correlated 0.664 with AI Overview visibility. Branded anchors correlated 0.527. Branded search volume correlated 0.392. Backlinks — the metric most agencies have been billing for since 2010 — correlated just 0.218.
Independent research from Kevin Indig at Growth Memo measured brand search volume specifically against ChatGPT visibility and found a 0.334 correlation. Seer Interactive’s parallel SERP-factor analysis put plain backlinks at just 0.10 against LLM mentions. A separate analysis of 75,000 brands by the Entities.org research team found brand mentions correlating r ≈ 0.66 versus backlinks at r ≈ 0.10.
Three independent measurement approaches. Three different sample sets. Three different methodologies. Same conclusion: brand signals predict AI citations far better than backlinks do.
That’s the headline. The rest of this guide unpacks what those numbers actually mean for link builders, which signals you should be investing in, and the specific tactics that move both metrics at once — because the brands winning in 2026 are not picking one over the other.
The correlation data, side by side
Here’s every published 2025–2026 correlation figure tracking off-site signals against AI visibility, consolidated. This is the cleanest cross-study reference available in the public domain.
| Signal | Correlation | Platform measured | Study |
| Branded web mentions | 0.664 | AI Overviews | Ahrefs, 75K brands |
| YouTube mentions | ~0.737 | ChatGPT, AI Mode, AI Overviews | Ahrefs, December 2025 |
| Branded anchors | 0.527 | AI Overviews | Ahrefs, 75K brands |
| Branded search volume | 0.392 | AI Overviews | Ahrefs, 75K brands |
| Brand search volume | 0.334 | ChatGPT | Kevin Indig, Growth Memo |
| Domain Rating (DR) | 0.25 | LLM mentions | Seer Interactive |
| Backlinks (total) | 0.218 | AI Overviews | Ahrefs, 75K brands |
| Branded ad traffic | 0.216 | AI Overviews | Ahrefs, 75K brands |
| Branded ad cost | 0.215 | AI Overviews | Ahrefs, 75K brands |
| Content volume | ~0.194 | ChatGPT, AI Mode, AI Overviews | Entities.org |
| Backlinks (plain) | 0.10 | LLM mentions | Seer Interactive |
| Backlinks | ~0.10 | ChatGPT, AI Mode, AI Overviews | Entities.org |
Two patterns jump out. First, every single brand-related signal correlates more strongly with AI visibility than every single backlink-related signal. Second, the highest-correlating signal of all isn’t even a brand mention — it’s a YouTube mention, which in Ahrefs’ December 2025 update measured r ≈ 0.737 across ChatGPT, AI Mode, and AI Overviews combined.
What “correlation 0.664” actually means in practice
Correlation coefficients run from -1 to +1. A 0.0 score means no relationship. A 1.0 score means a perfect linear relationship. Anything above 0.5 in observational SEO research is unusually strong. Anything below 0.2 is borderline noise.
Ahrefs’ own analyst team has been careful to note that 0.664 is still a moderate correlation in absolute terms — it’s not a deterministic rule. But it sits in the company of Google Keywords (r ≈ 0.65 to Google organic rank in Seer’s parallel work), which is the strongest single SERP-factor correlation in classic SEO research. In other words, brand mentions correlate with AI Overview visibility about as strongly as ranking keywords correlate with classic Google rank. That’s a meaningful comparison.
Backlinks at 0.218 sit in the same correlation range as branded ad traffic (0.216) and branded ad cost (0.215). The honest read of the Ahrefs numbers is: backlinks barely outperform paid advertising spend as a predictor of AI Overview visibility, and both of them sit far behind organic brand mentions.
Caveat — and this is important: correlation is not causation. Brands with strong mention profiles also tend to be older, more profitable, more aggressive about PR, and more entrenched in their categories. The mention itself may not be doing the work. But the directional signal is consistent across every published study, and it points the same way.
Why brand signals beat backlinks for AI citations
The architecture explains the data. AI systems are not search engines that index and rank pages. They are retrieval-augmented language models that assemble answers from training data plus real-time fetched content. Both phases reward different signals than classic PageRank does.
Training data favours frequency, not link graphs
During training, an LLM ingests billions of pages and learns statistical associations between entities. A brand that appears thousands of times across editorial coverage, forum discussions, news articles, and YouTube transcripts gets baked into the model’s prior probabilities. A brand that has 50 backlinks pointing at it — but barely any mentions across the wider web — does not.
Backlinks are a directed edge in a graph. Mentions are a frequency count in a corpus. AI training cares far more about the second.
Retrieval favours corroboration, not authority transfer
At inference time, modern AI systems cross-verify claims against multiple sources before citing anything. A backlink does not corroborate; it just points. A brand mention in a Tier-1 publication does corroborate — it’s a third-party signal that an editor judged this brand worth naming. Multiply that across 50+ outlets and you have what AI systems describe internally as an entity consensus signal: many independent sources naming the same entity in the same context.
Backlinks can contribute to that signal when they come with branded anchor text (0.527 correlation in Ahrefs’ data). But a plain backlink with a generic anchor like “click here” or a money keyword like “best CRM” contributes almost nothing to entity corroboration.
Distribution matters more than ownership
Stacker and Scrunch ran a controlled study across five leading LLMs and found that distributing content through third-party news outlets produced a 239% median lift in AI search visibility, with some test cases reaching 325%. The same content sitting only on the brand’s own site got nothing close. AI systems weight the authority of the domain doing the citing, not just the content being cited.
This is why a single Forbes mention can outperform 30 directory backlinks on AI visibility metrics. The Forbes mention adds entity consensus to the model. The directory backlinks add nothing the model cares about.
What this changes about link building in 2026
Backlinks are not dead. They still drive classic Google rankings, still influence DR/UR, and still feed referral traffic. The question is whether agencies still bill against backlink volume as the primary KPI, or whether they shift to billing against branded mention volume and entity coverage.
Most UK agencies running 2026 retainers are starting to make exactly this shift. The KPIs are evolving. Here’s how the metric mix actually looks in current best-practice retainers:
- Unlinked mentions tracked alongside linked mentions. Ahrefs’ Brand Radar now surfaces both. For AI citation visibility, an unlinked mention from BBC carries comparable weight to a followed link from a DR 50 trade publication.
- Referring domain diversity over backlink volume. Ten mentions across ten outlets produce more entity consensus than 50 mentions across two outlets. The new KPI is unique domains mentioning the brand per month, not backlinks acquired.
- Branded anchor text prioritised over keyword-rich anchors. Branded anchors correlated 0.527 with AI visibility — second-highest signal in the Ahrefs study. Exact-match anchors did not even make the top five. For AI-first link building, brand name in the anchor matters more than the target keyword.
- YouTube and Reddit footprint as standing line items. YouTube mentions correlated at 0.737 — the single highest signal in any published study. Reddit is Gemini’s #1 cited domain. Earned video coverage and credible Reddit footprint now belong inside the link building budget, not adjacent to it.
- Branded search volume as a leading indicator. Branded search volume correlated 0.392 with AI visibility. It’s a downstream metric — but it’s also a fairly clean proxy for whether the broader awareness machine is working. Track monthly branded search trend lines as an early-warning signal for AI visibility shifts.
If you’re still building out tactical infrastructure, our broader breakdown of the 15 link building strategies that work in 2026 covers which tactics produce both backlinks and mentions in the same workflow. The most efficient 2026 link building motions hit both KPIs simultaneously — digital PR, original research, and listicle placements all qualify.
Tactics that move brand signals (not just backlinks)
If you’re shifting your retainer scope from “acquire backlinks” to “acquire brand signals,” these are the highest-yield tactics ranked by the correlation they’re targeting.
1. Digital PR with named mentions
Pitching journalists to write stories that name your brand — with or without a link — is now the highest-value link building activity for AI citation lift. A named mention in BBC, Reuters, The Guardian, or any Tier-1 UK outlet contributes more entity consensus than 20 followed backlinks from DR 30 trade sites.
The trade-off: digital PR is harder to scale, harder to bill in fixed units, and harder to guarantee. But the per-placement value-to-AI-visibility ratio is significantly higher than guest posting or directory submissions.
2. Listicle placements (“best X for Y”)
Listicles on category-authority sites are extracted near-verbatim by Gemini and ChatGPT during retrieval. Getting your brand named in a listicle is functionally equivalent to ranking on the listicle author’s page — except the listicle author’s domain authority is doing the heavy lifting on the AI citation pipeline.
For UK and Indian SaaS, fintech, and DTC brands, listicle inclusion now ranks materially higher than guest posting on the per-placement value curve. Pitch the listicle author, not the editor.
3. Original research and survey data
Publishing original research — survey data, benchmark studies, sector analyses — earns named mentions every time the data gets referenced. The compounding effect is significant: one well-promoted research piece can earn 30–80 named mentions across Tier-1, Tier-2, and trade outlets over six months, each one contributing to entity consensus.
This is the tactic with the highest mention-volume-per-hour-invested ratio of any link building activity, full stop.
4. Podcast and YouTube guest appearances
YouTube mentions correlated at r ≈ 0.737 — the highest single signal in any published 2026 study. Podcasts often appear on YouTube as well. One quality podcast appearance with named brand and product mentions in the transcript can outperform 50 guest posts on AI visibility metrics.
If your link building agency is not pitching podcast bookings in 2026, that’s a meaningful gap in scope.
5. Reddit, Quora, and community-led mentions
Reddit is Gemini’s #1 cited single domain. Quora still feeds ChatGPT’s training set substantially. Authentic community mentions — earned through genuine participation, expert AMAs, or community-led ambassadorship — feed AI training data in ways that no traditional link building does.
Caveat: this is hard to do well and trivially easy to do badly. The line between “earned community footprint” and “obvious astroturfing” is thin, and Reddit moderators are vigilant. Done badly, this destroys brand reputation. Done well, it’s one of the highest-yield activities available.
6. Wikipedia and Knowledge Graph entity work
Wikipedia is one of the strongest single entity-authentication signals available. Gemini’s pipeline resolves named entity references to Knowledge Graph records before retrieval, and the Knowledge Graph leans on Wikipedia heavily. A verified Wikipedia entity, claimed Knowledge Panel, and Organization schema with sameAs links to authoritative profiles together form the entity-authentication baseline for any serious 2026 AI visibility programme.
This is slow, careful work. Wikipedia eligibility requires genuine third-party notability and the editors are strict. But it compounds permanently once achieved.
Ranked: AI citation lift per hour of effort
If you’re budgeting time across competing tactics, the order matters. Here’s the ranking based on observed citation lift per hour of effort across published 2026 case data:
| Rank | Tactic | Signal type | AI citation lift |
| 1 | Original research / survey data | Mention + link | Very high |
| 2 | Tier-1 digital PR (named mention) | Mention | Very high |
| 3 | Listicle placements on category-authority sites | Mention | High |
| 4 | Podcast / YouTube guest appearances | YouTube mention | High |
| 5 | Wikipedia / Knowledge Graph entity work | Entity authentication | High (long-lag) |
| 6 | Earned Reddit / Quora footprint | Community mention | Medium-high |
| 7 | Branded anchor text guest posts | Branded anchor + link | Medium |
| 8 | Trade press placements | Mention + link | Medium |
| 9 | Plain backlinks via outreach | Link only | Low-medium |
| 10 | Directory submissions | Link only | Negligible |
The pattern is clear: tactics that generate brand mentions outperform tactics that generate only backlinks on AI citation metrics, even when the link-only tactic produces a higher domain authority on paper.
Why you still need backlinks (despite the correlation gap)
The headline of this guide is that brand signals beat backlinks for AI citations. The full picture is more nuanced. Backlinks still matter in 2026 — they just do different work.
- Backlinks still drive classic Google rankings. And classic Google rankings still matter for AI Overviews. Among pages ranking #1 in Google, 43.2% were cited by ChatGPT according to Kevin Indig’s Growth Memo analysis — 3.5× higher than pages ranking beyond Google’s top 20.
- Backlinks still influence the Brave Search index. Brave is Claude’s retrieval backend, and Brave’s index ranks pages partly on classic link signals. No backlinks = no Brave visibility = no Claude citations on retrieval-grounded queries.
- Branded backlinks count as both signals. A backlink with branded anchor text contributes to both the backlink count and the branded anchor metric (0.527 correlation). One placement, two KPIs moved.
- Diversified linked mentions still produce trust signals. AI systems don’t ignore links. They just don’t weight unbranded volume the way Google’s PageRank did. Editorial links from authoritative outlets — the kind digital PR produces — count strongly in both worlds.
For the foundational picture of what link building is and why it still matters even with these shifts, see our breakdown of what link building means in 2026 and how it has evolved. The discipline hasn’t gone away. The metrics have.
How to measure brand signals for AI visibility
If you’re shifting your retainer KPIs toward brand signals, you need to actually measure them. Here’s the practical stack.
1. Branded search volume trend (monthly)
Track in Ahrefs Site Explorer under “Organic keywords by intent → branded.” Watch the monthly trend line. Rising branded search volume is the leading indicator for rising AI visibility. Flat or falling branded search volume despite link acquisition means the work is producing the wrong signal.
2. Unlinked brand mentions (monthly count and domain diversity)
Ahrefs Brand Radar, Mention.com, Brand24, and Meltwater all track this. The metrics that matter are total unique mention domains per month and the percentage that are net new (haven’t mentioned you before). Domain diversity, not raw volume.
3. YouTube mention count
Manual searches plus tools like Ahrefs YouTube Keywords. Track named brand mentions in video titles, descriptions, and (where you can extract transcripts) the actual video content. Any podcast or video appearance you do should be tracked here.
4. AI citation visibility (cross-platform)
Use Profound, Otterly, AthenaHQ, Quattr, or Generative Pulse by Muck Rack to track how often your brand is cited across ChatGPT, Gemini, AI Overviews, Claude, and Perplexity. Track citation breadth (number of prompts that cite you) rather than citation depth (where you appear in any one response).
Our broader review of the best link building tools available in 2026 covers the full stack across mention tracking, AI visibility, and traditional backlink analysis.
5. Wikipedia entity status and Knowledge Panel completeness
Binary check, but it matters more than any single tactical metric. Either you have a Wikipedia article and a claimed Knowledge Panel or you don’t. If you don’t, your entity authentication baseline is weak regardless of how strong your mention profile is.
Common mistakes when shifting to brand-signal link building
Most agencies and in-house teams make the same five mistakes when they try to pivot from backlink-only KPIs to brand-signal KPIs. Avoid these.
- Abandoning backlinks entirely. Backlinks still feed classic Google rank, Brave Search rank, and Claude retrieval. Don’t throw out the foundation to chase the new metric. Layer brand signals on top, don’t replace links with them.
- Buying unlinked mentions from low-authority sites. If a mention farm starts selling “AI visibility brand mentions” the way directory farms used to sell links, ignore them. AI systems weight mention authority, not mention volume. A mention from BBC outweighs 200 mentions from generic content farms.
- Skipping the entity authentication baseline. If your brand has no Wikipedia article, no claimed Knowledge Panel, and inconsistent NAP across the open web, brand mentions can’t compound the way they would for an entity-verified brand. Fix the baseline before scaling mentions.
- Treating linked and unlinked mentions as one metric. They do different work. Linked mentions feed both backlink-driven systems (Google, Brave) and entity systems. Unlinked mentions feed only entity systems. Track them separately, but report on both.
- Picking one platform’s correlation data as the whole truth. ChatGPT, Gemini, AI Overviews, and Claude weight signals differently. Ahrefs’ 0.664 figure is specifically for AI Overviews. The same signal correlates differently on Claude or Perplexity. Build for cross-platform consensus, not platform-specific optimisation.
The bottom line
Brand search volume, brand mentions, and branded anchors predict AI citations far more strongly than backlinks do. The Ahrefs 0.664 vs 0.218 figure is the most-cited data point in the AI search literature for a reason — it’s been replicated independently across three separate research methodologies.
But backlinks aren’t dead, and the brands winning in 2026 aren’t picking one signal over the other. They’re treating brand mentions as the headline KPI and backlinks as the underlying infrastructure that makes brand mentions earn their AI citation lift. Both metrics matter. The retainer scope just needs to track both.
If you’re updating your own measurement stack — or pitching a 2026 retainer to a client — pair this article with our running 2026 link building statistics. The full data set covers branded mention volume, AI citation rates, and backlink correlation figures from every major 2025–2026 study, kept current quarterly.
Build for brand signals. Don’t abandon backlinks. Measure both. That’s the whole game in 2026.
FAQ
Is brand search volume more important than backlinks for AI citations?
Yes, according to every published 2025–2026 correlation study. Branded web mentions correlate 0.664 with AI Overview visibility (Ahrefs, 75K brands). Backlinks correlate just 0.218. Independent measurement from Kevin Indig at Growth Memo and Seer Interactive replicates the directional finding: brand signals predict AI citations far more strongly than backlinks.
Should I stop building backlinks?
No. Backlinks still drive classic Google rankings, still influence Brave Search rank (which is Claude’s retrieval backend), and still feed referral traffic. Branded backlinks count as both signals — they contribute to both the backlink count and the branded anchor metric. The shift is from “link volume as primary KPI” to “brand mentions as primary KPI, backlinks as infrastructure.”
What’s the difference between a brand mention and a backlink?
A backlink is a clickable hyperlink from another site to yours. A brand mention is any reference to your brand name on another site, with or without a link. AI systems weight unlinked brand mentions almost as strongly as linked ones — unlike Google’s PageRank, which requires the link to transfer authority. Modern AI visibility programmes track both.
Why does YouTube correlate so strongly with AI visibility?
Ahrefs’ December 2025 update measured YouTube mentions at r ≈ 0.737 across ChatGPT, AI Mode, and AI Overviews — the highest single correlation in any published study. The likely reasons: YouTube transcripts are heavily ingested into AI training sets, YouTube is one of the few sources Gemini routinely cites as a primary source, and brands mentioned in video content tend to be brands with broader cross-channel footprint.
How do I increase branded search volume?
Branded search volume is downstream of brand awareness. It rises when more people hear about, read about, or discuss your brand. The activities that move it: PR coverage in publications your audience reads, podcast and YouTube appearances, paid awareness campaigns with strong creative, community-led ambassadorship, and content that names the brand explicitly in titles and lead paragraphs across the web.
Is correlation the same as causation here?
No, and Ahrefs’ own analyst team flagged this explicitly. Brands with strong mention profiles tend to be older, more profitable, more PR-active, and more entrenched in their categories. The mention itself may not be doing all the work — the brand strength behind it may be. But the directional signal is consistent across three independent research methodologies, and the practical implication is the same: invest in tactics that produce mentions, not just links.
Do unlinked mentions really count for AI citations?
Yes. Multiple 2026 studies confirm this. Ahrefs’ Brand Radar separates linked and unlinked mentions and finds both contribute to AI Overview visibility. The underlying reason: AI training corpuses are flat text, not link graphs. A brand name appearing in an article gets ingested into training data whether or not there’s a hyperlink attached to it.
What’s the single highest-yield tactic for moving brand signals?
Original research and survey data publication, by citation-lift-per-hour-invested. One well-promoted research piece can earn 30–80 named mentions across Tier-1, Tier-2, and trade outlets over six months, each one contributing to entity consensus. No other tactic produces the same mention-volume-per-hour ratio.
Does Wikipedia matter for AI citations?
Yes, especially for Gemini. Gemini’s pipeline resolves named entity references to Google Knowledge Graph records before retrieval, and the Knowledge Graph leans heavily on Wikipedia. Brands with a verified Wikipedia entity and a claimed Knowledge Panel are over-represented in Gemini outputs on entity-related queries. ChatGPT and Claude also use Wikipedia heavily — Wikipedia is ChatGPT’s #1 cited domain at 26-48% of its top-10 citation share.
Should small UK businesses worry about brand signals if they have no PR budget?
Yes — and small businesses actually have an advantage. Kevin Indig’s research found AI citation visibility in ChatGPT is less concentrated than classic search rankings, especially for niche topics. Small brands can win citation visibility in specific verticals through focused tactics: niche trade publications, podcast guest appearances, original research in a narrow domain, and authentic Reddit footprint. The headline KPIs scale, but the entry threshold is lower than it used to be for classic SEO.
