Backlinks and AI Citations Statistics for Private Practices (2026)

Do backlinks still matter for a private practice website?

Backlinks are still one of the clearest correlations in public SEO research, but the story has a second half most marketing content leaves out: most of the web has almost none, and the ones that do exist decay faster than you would expect.

Backlinko, with Ahrefs as a data partner, analyzed 11.8 million Google search results and found that pages ranking #1 have on average 3.8 times more backlinks and 3 times more referring domains than pages sitting in positions two through ten. That pattern has held up for years.

The honest caveat is that this is a correlation, not a controlled experiment, and the data is heavily skewed because most pages on the internet have few or no backlinks at all. In an earlier Ahrefs analysis of roughly 920 million pages, 55% had no referring domains whatsoever. Most SEOs still find it close to impossible to rank competitively for a contested keyword without any links, even though links alone never guarantee a ranking.

What backlinks cost, and why they don’t last

Ahrefs studied 450 sites that sell links and found the average purchased niche edit, a link added into existing content, costs $361.44, while a guest post placement averages $77.80, not counting the labor and outreach time behind either one. Practitioner estimates for digital PR links, the kind earned through original data or press coverage, typically run well over $1,000 per link once staff time is included.

Links also expire. Ahrefs tracked more than 2 million domains since 2013 and found at least 66.5% of the links pointing to them are now dead. A 2026 survey of 518 link building professionals by Editorial.Link found digital PR is now rated the single most effective tactic, at 48.6%, well ahead of guest posting and other approaches, though the same survey notes digital PR only works when there is a genuinely link-worthy asset behind it.

The content formats that earn the most links

Backlinko and BuzzSumo analyzed 912 million blog posts and found long-form content of 3,000 words or more earned 77.2% more referring domains than shorter posts. The same analysis found original research and data-driven content consistently outperformed opinion pieces and generic how-to guides for links, because journalists and other writers need a number to cite, not just an opinion.

This does not mean every post needs to hit 3,000 words. It means depth and original data correlate with links, while thin, generic content rarely earns anything, no matter how it is formatted. A single well-sourced statistics page built around your specialty or local market can out-earn a dozen generic listicles, which is exactly the bet this article is built on.

Local citations still matter for healthcare practices

Whitespark’s 2026 Local Search Ranking Factors report, based on a survey of local SEO experts, found Google Business Profile signals still carry the most weight in the local map pack, at roughly 32%. Citations, directory listings and mentions of your practice’s name, address, and phone number, are back in importance too. Whitespark introduced a new “AI Search Visibility” category in this year’s report, and citations account for a meaningful share of it.

For a therapy, dental, or veterinary practice, this means the old local SEO checklist has not gone away. A complete Google Business Profile, consistent name, address, and phone information across the directories you actually appear in, and steady, recent reviews are still foundational. They now double as signals AI systems draw on when answering a question like “therapist near me” or “best vet in [city].”

How ChatGPT, Perplexity, and AI Overviews choose what to cite

Profound analyzed 680 million AI citations between August 2024 and June 2025 and found Wikipedia is ChatGPT’s single most-cited source, accounting for 7.8% of all citations. A separate analysis of 30 million sources by Peec AI found Reddit ranked first or second across every major AI platform tested, including ChatGPT, Google AI Mode, Gemini, Perplexity, and Google AI Overviews, making it the most-cited domain overall.

The pattern across both studies is consistent. AI systems lean heavily on reference sites, community platforms, and a small number of authoritative publishers, not on the long tail of small business and practice websites. That does not mean a practice website cannot be cited. It means a generic page competing against Wikipedia and Reddit for a broad question is unlikely to win, while a page answering a narrow, specific question your ideal client is actually asking has a real shot.

The shrinking overlap between AI citations and Google rankings

In July 2025, Ahrefs found that about 76% of the sources cited in Google’s AI Overviews also ranked in the traditional top 10 for the same search. By March 2026, a larger Ahrefs study of 863,000 keywords and 4 million AI Overview URLs found that overlap had fallen to just 37.9%, meaning roughly six out of ten AI Overview citations now come from outside page one entirely.

BrightEdge’s separate research puts top-10 overlap even lower, closer to 17%, though its methodology differs from Ahrefs’ and the two studies are not directly comparable. Both agree on the direction. Google’s AI Overviews increasingly pull from a wider pool of pages than the ones ranking highest, which means holding the number one spot no longer guarantees an AI citation, and a page ranking on page two or three is not automatically out of the running.

The AI Overview click-through-rate drop

Seer Interactive tracked 3,119 informational search queries across 42 organizations, covering 25.1 million organic impressions, and found organic click-through rate dropped from 1.76% to 0.61%, a 61% decline, once an AI Overview appeared on the results page. Paid click-through rate fell even further, by 68%, over the same period.

Pew Research Center reached a similar conclusion from a different angle. Tracking nearly 69,000 searches from 900 US adults, Pew found people clicked a traditional search result only 8% of the time when an AI summary appeared, compared to 15% of the time when it did not, and clicked a link inside the AI summary itself in just 1% of visits.

Brand mentions: the signal that now beats backlinks

Ahrefs studied 75,000 brands in December 2025 and found that unlinked web mentions of a brand’s name correlate more strongly with AI Overview visibility than backlinks do, by a wide margin. In plain terms, being talked about by name across the web now matters more to AI visibility than being linked to.

This is a real shift from traditional SEO, where a mention without a link was often treated as close to worthless. For a private practice, it reframes digital PR, press coverage, podcast appearances, and even being named in a “best therapists in [city]” roundup article as AI-visibility work, not just brand-awareness work, whether or not any of it includes a clickable link back to your site.

AI referral traffic is tiny, but it converts far better

Semrush’s Traffic Channel Mix study, covering 17 industries across all of 2025, found AI referral traffic grew 66% over the year and outpaced every other channel, yet still made up under 0.15% of total website visits by year’s end. AI search is growing fast, from a very small base.

What makes that small slice worth paying attention to is quality. A separate Semrush study found the average visitor arriving from AI search converts at 4.4 times the rate of the average organic visitor. The likely explanation is that the AI has effectively pre-qualified the visitor by recommending your practice before they ever click, so the person landing on your site has already decided you are worth a look.

Does llms.txt actually help AI visibility?

llms.txt is a proposed file, similar to robots.txt, meant to tell AI systems what a site is about. The evidence so far says it does not move the needle. Ahrefs studied 137,000 domains and found only 28% publish an llms.txt file at all, and of those that do, 97% received zero AI crawler requests for the file in the month studied.

SE Ranking ran a separate analysis of roughly 300,000 domains and found no measurable link between having an llms.txt file and how often a domain gets cited by AI systems. Google’s John Mueller has been blunt about why. He has said no major AI service checks for the file, and compared it directly to the old keywords meta tag, a claim a site makes about itself that nobody actually verifies against.

Backlink statistics and AI citation data: the numbers at a glance

Every headline statistic in this report, with its source, the year, and how strong it is for citation. Strong means a recent primary source with a sound method and a large sample. Moderate means reputable but older, smaller, or vendor-run with a clear method. Weak means the claim could not be traced to a verifiable primary source.

The 2026 backlink and AI citation checklist for your practice

Put together, the data points to one durable strategy rather than a list of tricks. Earn a small number of genuinely relevant links instead of buying links you cannot defend. Keep your Google Business Profile and directory citations accurate and current. Build content specific and original enough that AI systems and human readers alike have a reason to cite it by name. And do not spend time chasing unproven levers like llms.txt.

Frequently asked questions

Methodology and sources

This report compiles published statistics from primary and reputable secondary sources. Every figure is attributed with its publishing organization and year, and every source link below was checked to confirm it is live and that the figure actually appears on the page. Where a widely repeated claim could not be traced to a verifiable primary source, it was left out rather than cited on faith. Statistics are presented as published and are not traffic forecasts for any individual website.