Link Building & Backlink Statistics 2026: What the Data Says About Links, Decay and AI Visibility

Links have been the awkward subject of SEO for twenty years. Everyone agrees they matter. Almost nobody agrees how much, and the industry that sells them has an obvious interest in the answer.

What has changed going into 2026 is that the question is no longer just about Google rankings. It is about whether an investment in referring domains also buys visibility inside ChatGPT, Perplexity and AI Overviews, where the published correlations are weaker than link sellers would like.

Below are the studies that actually have numbers behind them: index-scale crawls from Ahrefs and Backlinko, survey data from uSERP and Aira, a 2025 Semrush correlation study on backlinks and AI mentions, and link rot research from the Pew Research Center. Where a figure is old, the year is stated.

Key link building and backlink statistics for 2026

• Roughly 95% of all pages have zero backlinks, according to Backlinko's analysis of 11.8 million Google search results [3].
• Ahrefs found that 66.5% of links pointing to the 2,062,173 websites it sampled had rotted over nine years, rising to 74.5% once temporary errors are counted [5].
• The Pew Research Center found that 38% of webpages that existed in 2013 were no longer accessible in 2023, against 8% of pages from 2023 [6].
• 74.3% of link builders buy backlinks, and paid links cost $83 on average, per an Authority Hacker survey of 755 link builders reported by Ahrefs [7].
• When Ahrefs pitched 450 sites across nine niches for paid link insertions, only 12.6% were willing to sell, at an average of $361.44 per niche edit [7].
• uSERP's survey of 800+ SEO professionals found 67.5% believe backlinks have a "big impact" on rankings, and 46.5% spend $5,000 to $10,000 a month on link building [8].
• Semrush measured a Pearson correlation of 0.65 between Authority Score and AI mentions across 1,000 domains, but only about 0.23 against AI Share of Voice [17].
• Only 8.5% of outreach emails get a response, from Backlinko's analysis of 12 million outreach emails [18].
• The correlation between the share of exact-match anchors and rankings is 0.1436 (Spearman, average), which Ahrefs described as relatively weak across 384,614 pages [13].
• 10.6% of all backlinks to the top 110,000 sites are nofollow, whilst rel="sponsored" appears on just 0.01% of referring domains, per Ahrefs [15].

Do backlinks still correlate with rankings in 2026?

Yes. What is disputed is direction and magnitude. The largest public studies show top-ranking pages hold more referring domains and gain them faster, but none isolate cause from effect.

Backlinko's study of 11.8 million search results found the number one result had 3.8x more backlinks than positions two to ten, and 3x more referring domains [3]. First published 2016, last updated April 2025, so treat it as a long-standing pattern rather than a fresh reading.

Ahrefs' study of 200,000 top-ranking pages across 10,000 keywords found number one pages gain followed backlinks from new referring domains at +5% to +14.5% per month, with the pace declining as ranking position declines [2]. The same researchers said they "failed to distill any observable relationship" between acquisition pace and subsequent SERP movement over three months.

That second finding is the one most link building pitches omit. A page can be gaining links because it ranks, not ranking because it gains links.

Google's own published position is deliberately soft. Its How Search Works documentation says only that "one of the factors used to determine quality is understanding if other prominent websites link or refer to the content" [12]. Note the phrase "link or refer". Google has quietly widened its own language beyond hyperlinks.

What share of pages have no backlinks at all?

Around 95% of pages on the web have no backlinks whatsoever, and the overwhelming majority of those pages get no search traffic either. The distribution is not merely skewed. It is close to degenerate.

Backlinko's crawl put the figure at approximately 95% of all pages with zero backlinks [3]. Ahrefs' index tells the same story from the other direction: its Content Explorer index holds roughly 20 million pages with no referring domains, of which only 2,997 get more than 1,000 search visits a month [4]. About one in 6,671.

The practical reading is less dramatic than it sounds. The zero-backlink majority is dominated by pagination, tag archives, thin ecommerce variants and abandoned blogs, not by pages anybody was seriously trying to rank. A site needs enough domain-level authority and internal linking that individual pages inherit strength, not a link to every URL.

How much does a backlink cost in 2026?

There is no single price. Reported averages run from $83 to well over $1,000 depending on whether you ask link buyers what they paid, ask agencies what they charge, or email site owners and see what they quote.

uSERP's survey found monthly budgets clustering at $5,000 to $10,000 for 46.5% of respondents and $1,000 to $5,000 for 35.5% [8]. Ahrefs reports SEO-focused digital PR agencies typically hold minimum retainers of $4,000 to $15,000 a month [10].

The $83 average and the $1,500 premium tier do not contradict each other. They measure different populations. Surveys of link builders capture a long tail of cheap placements bought in volume from marketplaces. Agency rate cards capture the few genuinely hard-to-get placements a client is billed for individually.

If you are budgeting, use the agency figures. The $83 average mostly tells you that most bought links are cheap because they are worth very little.

How many SEOs buy links?

Most of them. Ahrefs reports 74.3% of link builders buy backlinks despite it being an explicit violation of Google's policies [7]. That is the most awkward statistic in the discipline.

Google's spam policies, last updated May 2026, define link spam as "the practice of creating links to or from a site primarily for the purpose of manipulating search rankings", listing "exchanging money for links, or posts that contain links" among the examples [11].

The supply side is more selective than you might expect. When Ahrefs emailed 450 sites across nine competitive niches asking to buy link placements, only 12.6% were willing to sell [7]. On the guest posting side, 25.5% of 180 sites responded to a pitch, and roughly half of those who responded asked for a fee unprompted, at an average of $77.80 [7].

That 12.6% is the number worth sitting with. If you assume every site is quietly for sale, you will burn a lot of outreach hours discovering otherwise. The paid link market is real and widespread, but concentrated in a specific segment of the web rather than universal.

Note how little of the web discloses. Ahrefs found rel="sponsored" on only 0.01% of referring domains to the top 110,000 sites, and rel="ugc" on 0.44% [15]. Six years after Google introduced the attributes, near enough nobody uses them.

How fast do backlinks decay?

Fast. Link building is less an accumulating asset than a leaky bucket, and the leak rate is measurable.

Ahrefs' link rot study of 2,062,173 websites found that 66.5% of links pointing at them over the previous nine years had rotted, rising to 74.5% once temporary errors and other SEO-relevant issues were included [5]. That study was published in April 2022 and updated in February 2024.

Pew Research Center's independent work reaches the same conclusion from a different angle:

Source: Pew Research Center, sampling nearly one million pages from Common Crawl plus 500,000 government pages, 500,000 news pages and 50,000 Wikipedia pages [6].

For anyone modelling link building ROI, this matters more than acquisition cost. A £500 link that survives eight years and a £500 link that survives eighteen months are entirely different purchases, and almost nobody prices them differently. It also explains why sites that stop link building do not hold position indefinitely. Standing still is a slow decline.

Does anchor text still move rankings?

Less than the folklore suggests externally, more than expected internally. The correlation data on external anchors is weak enough to be uncomfortable for anyone selling exact-match placements.

Ahrefs analysed 384,614 pages across 19,840 keywords and found the correlation between the percentage of exact-match anchored links and rankings was 0.1436 (Spearman, average) and 0.1869 (median) [13]. Phrase-match anchors scored lower still at 0.1057 average. Ahrefs' own conclusion was "a relatively weak correlation". That study dates from 2019 and was updated in 2020, so it predates several core updates.

The internal linking picture is very different. Zyppy's study of 23 million internal links across 1,800 websites and roughly 520,000 URLs found that pages with at least one exact match anchor had at least five times more traffic than pages without [14]. The relationship between anchor variety and search clicks survived removing about half the outliers. Published July 2022, updated February 2026.

Anchor text is a relevance hint you fully control internally and barely control externally. Optimising the anchors you own is cheap and evidently productive. Chasing exact-match anchors from third parties is expensive, risky under Google's link spam policy, and supported by a correlation of about 0.14.

Worth noting alongside this: Ahrefs found 73.6% of 140,592 domains have reciprocal links, and 43.7% of 112,440 top-ranking pages carry some [16]. Reciprocal linking is normal web behaviour, not a red flag, though that data is from 2019.

Do backlinks help you get cited by AI search?

Partly. Authority correlates with being mentioned by AI systems, but the strongest published predictors are not links.

Semrush tracked 1,000 randomly selected domains across ChatGPT, ChatGPT with Search, Gemini, AI Overviews and Perplexity in a study published in October 2025 [17]. Authority Score against raw AI mentions produced a Pearson correlation of 0.65 and Spearman of 0.57. Against AI Share of Voice, the figures dropped to roughly 0.23 Pearson and 0.36 Spearman.

Interestingly, nofollow links correlated marginally higher with AI mentions (0.340 Pearson, 0.509 Spearman) than follow links (0.334 and 0.504) [17]. If links were passing authority into these systems the way PageRank passes it into Google, you would not expect that.

The likelier explanation is that link counts are a proxy for brand size, and brand size is what the models respond to. Ahrefs' July 2026 review of over three million US queries found the most cited AI Overview domains were YouTube at 21.1%, Reddit at 18.5% and Facebook at 10.7%, a platform-share picture rather than a link-equity one [21]. Its study of 1.4 million ChatGPT 5.2 prompts found 88.46% of citations came from the search retrieval path, and centred its analysis on semantic similarity between prompt and page title rather than any link metric [20].

So links are not useless for AI visibility. They help you rank, ranking feeds retrieval, and retrieval feeds citation. But a link budget aimed at AI citations is a very indirect instrument.

What response rate should you expect from outreach?

Around 8.5%, and that is a reply of any kind, not a link. Anyone quoting a 30% success rate is counting something else.

Backlinko's analysis of 12 million outreach emails found an 8.5% response rate, with a single follow-up boosting replies by 65.8% and personalised subject lines lifting response by 30.5% [18]. Published in 2019, and email norms have hardened since, so treat 8.5% as a ceiling rather than a floor.

Aira's survey of 270 industry professionals found 32% take one to two hours to secure a single link and 24% take three to five hours [19]. That is a labour figure, not a link figure. Multiply it against a UK agency hourly rate and the $83 average paid link looks less like a shortcut and more like an inevitability.

The same data found 80% of respondents believe brand mentions influence organic rankings [19]. The industry was drifting towards mentions before AI search made it fashionable.

How to read these numbers

Almost every link study here is correlational, run on a crawler's index rather than Google's, and unable to separate cause from effect. Pages that rank attract links. That alone is enough to generate every ranking correlation in this article without links causing anything.

Sample bases differ wildly. Ahrefs' Content Explorer index of 14 billion pages is explicitly biased towards higher-quality content compared with its 340 billion page Site Explorer index [4], so "share of pages with no backlinks" means something different depending on which index you use.

Survey data is self-reported and self-selecting. uSERP's 800+ respondents and Aira's 270 opted into a link building survey, so they are more link-invested than the average marketer.

Several key figures are old, including several Ahrefs still carries in its own SEO statistics roundup, last updated June 2026 [1]. The Backlinko ranking study dates from 2016, the anchor study from 2019, the backlink growth study from 2018. They are cited because nothing larger has replaced them, not because they are current.

Pricing figures are the least reliable of all. Nobody publishes an audited average price of a backlink, because the market is informal, undisclosed and full of participants with a reason to inflate or deflate their answer.

What this means going into 2027

Budget for decay, not just acquisition. With two-thirds of links rotting over nine years [5] and 38% of a 2013 page cohort gone by 2023 [6], model a replacement rate into any link building business case.

Expect measurement to shift from links to mentions. Google's own wording is now "link or refer" [12], 80% of Aira respondents believe mentions influence rankings [19], and the AI data points the same way [17]. Reporting that counts only referring domains will look dated.

Cheap paid links will keep losing value faster than they lose popularity. 74.3% of link builders buy them [7] at an $83 average, which is exactly the price point that signals a link nobody would have given away.

Internal anchors remain the most underused lever. A five-times traffic difference from exact-match internal anchors [14] against a 0.14 correlation for external ones [13] is not a close call, and internal linking costs nothing but attention.

Treat AI visibility as a brand problem with a link component, not the reverse. Published correlations put authority at 0.65 against raw mentions but 0.23 against share of voice [17]. That gap is where the 2027 budget argument gets fought. If you want help working out how findable your business really is in search and AI answers, you can book a call with me.