You’re paying for links, and you can’t tell if it’s working.
That’s the real question underneath “do backlinks still matter in 2026?” — and it’s why this query’s search results are a mess. Search it yourself: Reddit threads and Quora answers hold the top spots, thin agency posts fill the middle, and one of the highest-ranking reference pages confidently declares backlinks “a top-3 Google ranking factor” — a claim Google’s own search analysts have publicly contradicted, on stage, more than once.
So the person asking is stuck between three unreliable narrators. The link-selling industry, whose answer is always yes, because their invoice depends on it. The “SEO is dead” crowd, whose answer is always no, because their content depends on it. And Google, whose answers are honest but scattered across a decade of conference remarks nobody has bothered to assemble.
This post assembles them. The full on-record history of what Google has actually said about links. What links still verifiably do. What the 2026 data shows about how the trust signal is changing — including numbers that should genuinely change how you spend. And the reframe that resolves the whole tired debate: the question was never really about links.
Short version, so you have it up front: yes, backlinks still matter in 2026 — but their job has changed, their relative weight has fallen, and the thing they were always standing in for has become more valuable than the links themselves. If you’re buying links by the bundle, you’re buying the wrong thing. Here’s the evidence.
Why links mattered in the first place
To re-price something, you have to know what it was originally pricing.
A backlink is a hyperlink from one website to another. Google’s founding insight — the idea behind PageRank, the algorithm the company was built on — was that a link is a vote. When one site links to another, it’s vouching: this is worth your time, this is the source, go look. Count the votes, weight them by the credibility of who’s voting, and you get a machine-readable map of reputation. It worked spectacularly, because in the late-90s web, a link was the only scalable, public signal of one entity trusting another.
Hold onto that framing, because everything else in this post hangs on it: the link was never valuable in itself. It was valuable as evidence — a recorded act of one party vouching for another. Google didn’t want links. Google wanted reputation, and links were the best proxy the web offered.
Two things follow, and both predicted the last decade before it happened.
First: any proxy that becomes a currency gets counterfeited. The moment links moved rankings, an industry emerged to manufacture them — link farms, paid placements, exchanges, private blog networks. Every counterfeit devalued the signal, and Google has spent twenty years in an arms race to discount the fakes.
Second: a proxy is only irreplaceable until better evidence arrives. If Google ever developed other reliable ways to read reputation and content quality — hundreds of other signals, language models that assess content directly, entity graphs that track who’s mentioned where — the weight on the original proxy would naturally fall. Not to zero. Just toward its honest value.
Both of those things happened. Here’s the record.
What Google has actually said: a decade of statements
The industry treats “links matter less now” as a hot take. It’s not. It’s the most consistently repeated message Google’s own people have delivered for over a decade — the assembled record reads like one long memo nobody opened.
| Year | Who | What they said |
| 2014 | Matt Cutts, then head of webspam | Backlinks “still have many, many years left in them” — but over time, as Google gets better at understanding content quality directly, they’ll become “a little less important” |
| ~2018 | John Mueller, Search Advocate | The weight on links will “drop off a little bit” over time; links will always matter somewhat, because Google has to discover pages somehow — but “it won’t be such a big factor as sometimes it is today” |
| 2020 | John Mueller | Links are “definitely not the most important SEO factor” |
| 2022 | Duy Nguyen, Search Quality team | Backlinks as a signal have “a lot less significant impact compared to when Google Search first started out” — Google now has hundreds of robust ranking signals |
| Sept 2023 | Gary Illyes, Search Analyst (Pubcon) | Links are important, “but I think people overestimate the importance of links. I don’t agree it’s in the top three. It hasn’t been for some time” |
| 2024 | Gary Illyes | Google needs “very few links” to rank pages; over the years links have deliberately been made less important |
The 2023 and 2024 remarks are documented by Search Engine Land and Search Engine Journal respectively, and the earlier statements are compiled in the same coverage. Illyes even offered a concrete example at Pubcon: a site with excellent content that ranked #1 after being discovered through nothing more than its sitemap. No link profile at all.
Notice three things about this record.
It’s directional, not binary. Nobody at Google has ever said links don’t matter. Every statement says the same thing: important, less than before, less than you think, trending down. The honest answer to “do backlinks matter?” has been “yes, and less each year” since before most current link-building packages were designed.
It’s a ten-year-old message. Cutts predicted this trajectory in 2014. Everything since has confirmed it. An industry that still sells links as the ranking lever isn’t behind the news — it’s behind a decade of news.
And the market ignored it, for a predictable reason. Links are the most sellable deliverable in SEO. A link is countable, screenshot-able, invoice-able. “We built you 40 links this month” fits a report in a way “we made your site more genuinely trusted” never will. The industry kept selling links at 2014 prices because the product was convenient — not because the signal held its value. When you see a page ranking for this very query claiming links are still a top-3 factor, against the direct on-record statement of Google’s own search analyst, you’re watching that incentive in action.
What links still do — this part is not dead
Re-pricing is not writing off. Strip away the inflation and links still do real, verifiable work in 2026 — and pretending otherwise is the “SEO is dead” crowd’s version of the same dishonesty.
Discovery still runs on links. Mueller’s caveat in that table is the permanent floor: Google finds pages by following references to them. A page nothing links to — internally or externally — is a page the crawler has to be told about. Links are the web’s circulatory system before they’re a ranking input, and that function isn’t going anywhere.
A genuine link is still the clearest single act of vouching. When a respected industry publication, a professional body, or a well-known specialist links to your work, that’s a public, verifiable, costly-to-fake endorsement. Google has repeatedly pointed to links from authoritative, relevant sources as part of how quality is assessed — the arms race was always against fake votes, not voting.
Quality-link profiles still correlate with winning — the distribution just sharpened. What’s changed inside the signal is the value curve. Relevance and source credibility now do nearly all the work: a handful of genuine references from respected sites in your actual field outweighs hundreds of directory listings, reciprocal swaps, and guest-post-farm placements — which round to zero, or worse, cost trust. The old game was volume. The current game is: would this linking site vouch for you if no SEO were involved?
And the floor for needing them has dropped. Illyes’s “very few links” remark cuts both ways. It deflates the link-selling pitch — and it liberates smaller, focused sites. In specific verticals and local markets, precise content, clean structure, and a coherent entity can rank with a link profile that would have been considered hopeless a decade ago. If you’ve been told you can’t compete until you’ve “built authority” through a year of paid placements, you’ve been quoted 2014 prices for a 2026 market.
Links are E-E-A-T’s oldest receipt
Here’s where the backlink question connects to the framework Google actually uses to talk about quality — and why the two conversations are really one.
E-E-A-T — Experience, Expertise, Authoritativeness, Trustworthiness — is Google’s vocabulary for content and publisher credibility, drawn from its Search Quality Rater Guidelines. Look at where links live inside that framework and their real, durable role snaps into focus.
Experience and expertise are things you demonstrate on your own pages — first-hand evidence, credentials, depth. Authoritativeness is different: it’s the one pillar you cannot self-declare. Authority exists only in what the rest of the web says about you — and a link is the oldest machine-readable form of the rest of the web saying something about you. When Google’s Gary Illyes was asked years ago how E-A-T is assessed, his reported answer was that it’s largely based on links and mentions on authoritative sites. The rater guidelines themselves instruct human raters to research a site’s reputation across the wider web — links, mentions, reviews, coverage.
So place the backlink correctly in the hierarchy: a link is one receipt in your reputation file. A strong one — public, weighted by who issued it, hard to fake credibly. But a receipt nonetheless. The file is what Google is actually reading, and the file has always contained more than hyperlinks: unlinked mentions, reviews, coverage, the professional footprint of your authors, the consistency of your entity everywhere it appears. (We’ve written a full guide to that file — what E-E-A-T actually is and how to build it — and this post is effectively the chapter on its most over-sold line item.)
This is also why buying links never made sense even when it “worked”: you were adding forged receipts to a file that gets audited. It’s why a link from one respected publication in your field outweighs fifty from nowhere: the receipt is only worth the reputation of whoever signed it. And it’s why what happens next in this post shouldn’t surprise you — because if the file matters more than any single receipt, then as Google’s systems got better at reading the whole file, the hyperlink’s exclusive claim on reputation had to end.
In 2025–26, it did — measurably.
The 2026 twist: the vouching signal outgrew the hyperlink
Two datasets published in the last year quantify the shift, and both come from Ahrefs — a company whose core product is, of all things, a backlink index. When the link-data company’s own research demotes link metrics, it’s worth reading carefully.
First: what correlates with AI visibility. Ahrefs studied 75,000 brands to find which factors correlate with being mentioned across ChatGPT, Google AI Overviews, and AI Mode. The results, by correlation strength: mentions on YouTube — in titles, transcripts, descriptions — led at roughly 0.74. Branded web mentions, linked or not, ran 0.66–0.71 across platforms. And Domain Rating — the classic distillation of backlink authority — trailed at 0.27–0.33 depending on platform. Link metrics sat mid-pack; being talked about dominated.
Sit with what that means. For visibility in the AI layer — the layer answering a fast-growing share of your prospects’ questions — the raw strength of your backlink profile correlates half as strongly as simply being mentioned across the web, and about a third as strongly as being present on video. Correlation isn’t causation, and Ahrefs is careful to say so. But the direction is consistent across every platform they measured, and it matches the mechanism: AI systems read reputation from the whole record — text, mentions, transcripts, entities — not from a link graph alone. The hyperlink was reputation’s format for twenty-five years. The format requirement is dissolving; the reputation requirement is not.
Second: what feeds the AI answers. Ahrefs’ AI Overview research tells the companion story. Their July 2025 study of 1.9 million citations found 76.1% of AI Overview citations ranked in Google’s top 10 for the query; their March 2026 re-run across 863,000 SERPs found that share had fallen to roughly 38%, as more aggressive query fan-out pulled citations from related sub-query SERPs instead of the head keyword’s. In the same update: YouTube is now the single most-cited domain in AI Overviews, making up 18.2% of citations that don’t rank in the top 100 at all.
Assemble the two findings and the strategic picture for 2026 is clear. Visibility is decided less by your position on one keyword (where concentrated link equity was the classic lever) and more by two things links only partially control: how thoroughly you cover a topic’s whole question-space, and how widely and consistently you’re vouched for across the open web — in text, in communities, on video. The link is still in that picture. It just stopped being the frame. (For the mechanics of how the engines actually choose who to cite, we’ve broken that down separately — this post is the “why the old currency devalued” half of that story.)
So the question was wrong
“Do backlinks still matter?” treats the link as the asset. Twenty years of Google statements, the E-E-A-T framework, and the 2026 correlation data all point at the same correction:
A backlink was never the asset. It was the receipt.
The asset is, and always was, being genuinely vouched for — by publications, practitioners, communities, and platforms that your market already trusts. The hyperlink was simply the first format that reputation came in that a machine could count. Google weighted the receipt heavily while it was the only evidence available, discounted it as counterfeits flooded in, and is now reading the rest of the file: unlinked mentions, reviews, video presence, entity consistency, the professional reality of the people behind the content.
That single reframe answers every version of the question this SERP is arguing about:
- “Do links still matter?” Yes — receipts are still evidence, and genuine ones from credible issuers are strong evidence.
- “Do they matter as much as before?” No — the auditor reads far more of the file now, and has said so for a decade.
- “Should I buy links?” You’re asking whether to buy forged receipts for an audit that’s getting better every year at spotting forgeries — and that increasingly doesn’t need receipts to see whether the underlying transactions happened.
- “How many backlinks do I need?” Wrong unit. The question is how many credible parties in your field would genuinely vouch for you — and whether there’s anything on your site worth vouching for.
The uncomfortable, liberating implication: link building was never really the discipline. Reference earning is. Publish the thing worth citing, be present where your field talks, and the receipts — linked and unlinked — accumulate as a side effect. Run it the other way, receipts first, and you’re decorating an empty file.
What to actually do — and what to demand from anyone you pay
Everything above collapses into five moves. None requires a big team. All compound.
1. Build assets worth referencing — this is the whole engine. The only link strategy that has survived every algorithm era is giving credible sites in your field a reason to reference you: original data no one else has, a genuinely useful tool or framework, the definitive answer to a question your industry keeps asking, a clearly argued position. One citable asset outperforms a year of manufactured placements, and it keeps earning after the invoice would have stopped.
2. Widen the vouching surface beyond hyperlinks. The correlation data says being mentioned now out-signals being linked for the AI layer — so work the whole file. Expert commentary in industry press. Podcast and webinar appearances. A real video presence, even a modest one, given YouTube’s outsized weight in AI citations. Reviews gathered and answered as an operating discipline. Genuine participation where your field already talks. Every one of these is reputation the systems can read, no anchor text required.
3. Judge link quality like an auditor, not an accountant. When links do come, count what matters: unique referring domains over raw totals, topical relevance over generic authority, and the one test that survives everything — would this site have referenced us if no SEO were involved? Prune the vanity metrics from your reporting; a rising count of worthless links is a cost dressed as a KPI.
4. Keep the plumbing honest. Links remain how crawlers move. Internal linking that routes authority to your money pages, clean sitemaps, no orphaned content — the unglamorous discovery layer still decides whether anything else gets seen. This is an afternoon of discipline per quarter, not a retainer.
5. Interrogate anyone selling you links. If you buy link building — and there are legitimate versions, usually sold as digital PR — make the vendor pass the questions this post equips you to ask: Show me the sites you’d target and why they’re relevant to my field, specifically. Would these sites reference me without payment, and if not, what makes the placement defensible? What’s the asset you’re pitching — or are you placing links to pages nobody would cite? How will you report quality, not counts? A vendor with good answers is selling reference earning. A vendor with volume tiers and a price per link is selling forged receipts, at 2014 prices, for an audit that’s already moved on.
We’ll say the quiet part plainly: we don’t sell link packages. Not as restraint — as consistency. An agency accountable for enquiries and booked appointments can’t invoice for a countable deliverable it knows is a proxy, and a discounted one at that. The work that actually moves the file — citable assets, earned mentions, entity consistency, topic-wide coverage — is just called doing the job.
Backlinks were never the goal. Being the practice, the school, or the firm your market already vouches for — that was always the goal, and it’s the one thing every algorithm era has rewarded. We build that as one connected search system: the organic foundation that earns the references, and accountability you can audit — enquiries, patients, students, and clients, never link counts.
Frequently Asked Questions
Are backlinks still a Google ranking factor in 2026?
Yes — but a diminished one, by Google’s own account. Google’s Gary Illyes stated on the record that links haven’t been a top-3 ranking factor “for some time,” and that Google needs “very few links” to rank pages. Links still aid discovery and still signal genuine endorsement; they no longer carry the dominant weight the link-selling industry prices them at.
How many backlinks do I need to rank?
There’s no number, and chasing one is the wrong frame. Google has confirmed sites can rank with very few links when content, structure, and relevance are strong — especially in specific verticals and local markets. What matters is whether credible, relevant sites in your field genuinely reference you, not how many domains of any quality do.
Do backlinks matter for ChatGPT and AI Overviews?
Less than being mentioned does. Ahrefs’ 75,000-brand study found Domain Rating correlated with AI visibility at roughly 0.27–0.33, while branded web mentions ran 0.66–0.71 and YouTube mentions led at about 0.74. AI systems read reputation from the whole web record — mentions, transcripts, entities — not the link graph alone. Links contribute; they no longer define.
Are paid backlinks worth it?
Paid placements are forged receipts: they imitate the vouching signal without the vouching. Google’s systems have spent two decades learning to discount exactly this, its spam policies target it explicitly, and the trend in every dataset is toward reading reputation more broadly — making bought links a depreciating, risk-carrying asset. Money spent manufacturing links buys more when spent creating something worth referencing.
Can a new website rank without backlinks?
Increasingly, yes — within limits. Google has cited examples of link-poor pages ranking on content strength alone, and its analysts say very few links are needed in many verticals. A new site with precise topical coverage, clean structure, and a consistent entity can win specific queries early. Competitive head terms in contested markets still favour established reputations — which is an argument for starting narrow, not for buying shortcuts.
