A business can publish a convincing explanation of its expertise on its own website. It is harder to persuade an industry publication to reference that explanation, a researcher to use its findings or a customer to recommend it publicly. Those references give readers something beyond the business’s own account of itself.
That is why backlinks remain worth working for. A backlink is simply a link from another website to yours, but the reason it exists matters. Someone may be pointing readers towards evidence, crediting a source or recommending a useful service. They may also have been paid to place it. Counting the links tells you very little about that difference.
AI search gives us another reason to examine these references. Recent studies associate stronger backlink profiles with more AI citations, while research on brand mentions points towards the importance of recognition across the wider web. For a business deciding where to invest, this makes earning credible references a useful part of its SEO work. It does not establish that buying more links will make an AI assistant recommend it.
What the research says about backlinks and AI search
There is now evidence worth discussing beyond the familiar argument that links have always mattered to Google. The studies below examine different outcomes, so their figures should be read within each study rather than combined into a single ranking of AI-search factors.
| Study | What it measured | Finding to keep in mind |
|---|---|---|
| SE Ranking November 2025 | ChatGPT citations across 129,000 domains | Referring domains were the strongest predictor in its model. This was not a link-building experiment. |
| Semrush October 2025 | Authority Score and visibility across five AI experiences | Association was stronger for mention counts than share of voice. Authority Score combines several measurements. |
| Ahrefs December 2025 | Brand mentions in three AI experiences, across 75,000 brands | Web mentions and branded anchors correlated more strongly than raw backlink metrics. |
Referring domains and ChatGPT citations
In November 2025, SE Ranking studied 129,000 domains and 216,524 pages, using responses to 100,000 prompts. Its predictive analysis identified referring domains as the strongest factor associated with citation frequency in its ChatGPT dataset.
Sites with up to 2,500 referring domains averaged 1.6–1.8 citations, compared with 8.4 for sites with more than 350,000. A referring domain is a separate website linking to yours; many links from one website still count as one referring domain.
This is evidence of an association between established link profiles and citation frequency. It is not a target of 350,000 domains for a smaller business, or a forecast of what acquiring another link will achieve. The study used a predictive model, rather than an experiment that added links while holding everything else constant.
Authority scores and AI mentions
Semrush’s October 2025 study of 1,000 domains compared backlink-related measurements with visibility across five AI experiences. Authority Score and AI mentions had a Pearson correlation of 0.65 and a Spearman correlation of 0.57. These coefficients describe how the measurements vary together; they are not percentage increases in visibility.
The relationship was weaker when the outcome was AI share of voice: 0.23 and 0.36 respectively. Authority Score also combines several measurements, so this result cannot isolate the contribution of backlinks. It supports the case for established websites tending to appear more often, without showing which part of their wider reputation caused that appearance.
Brand references often show a stronger relationship than link volume
In December 2025, Ahrefs compared 75,000 brands across ChatGPT, AI Mode and AI Overviews. Branded web mentions correlated with visibility at roughly 0.66–0.71. Branded anchor text, where a link names the brand, also showed positive correlations: 0.628 for AI Mode, 0.527 for AI Overviews and 0.511 for ChatGPT. Raw backlink metrics were much weaker.
Ahrefs measured whether brands appeared in answers, which differs from measuring whether a particular page was cited as a source. Its sample also selected domains with a Domain Rating above 40 and a highest-volume keyword with at least 800 monthly searches. The findings therefore need care when applied to a new or little-known business.
How to read the evidence. Together, these studies give us a reasonable basis for treating backlinks as part of building visibility in AI search. They also show why link totals alone are an inadequate goal. Large brands tend to have more links, more coverage and more useful pages, and observational research cannot neatly separate all those advantages.
A backlink, a mention and an AI citation do different jobs
A reader can follow the link to your report.
A trade publication uses a finding and links to the original report. You have earned a backlink. If it also names your business, you have earned a brand mention too.
What to record: The referring page, destination URL and relevant referral visits.
These outcomes can overlap. A backlink does not guarantee an AI citation, and a citation does not guarantee a brand recommendation or a visit.
This distinction matters when you commission PR or review an AI-visibility report. A brand recommendation, a source link and a visit to your website are separate results. Decide which would help the business and keep them separate in reporting.
A recent October 2026 preprint by Mahe Chen and Nick Koudas, covering 158,099 answers and 468,037 citations, also distinguishes source retrieval from brand knowledge. In their analysis, which pages ChatGPT retrieved explained much more variation in citations than page properties did. Their findings help explain why being retrieved as a source and being named as a brand need separate attention. The paper is a preprint and does not demonstrate a causal effect from backlinks.
Which backlinks are worth earning?
Start by reading the page that would contain the link. Who uses it? What does it help them understand? Would a reader have a reason to follow the reference to your website? These questions often tell you more about a potential placement than the domain’s headline score.
A small industry association might maintain a useful resource page for exactly the people you serve. A much larger publication might offer a paid article in a section that has little connection to its usual readership. The larger domain deserves no automatic preference.
The surrounding text matters as well. A link supporting a specific finding gives the reader a clear reason to visit the source. A company name dropped into an unrelated paragraph offers much less context. Descriptive anchor text is helpful when it belongs naturally in the sentence; prescribing the same keyword phrase for every placement is difficult to justify editorially.
Before you pursue a placement
- Read the page. Does the publisher cover the subject with care?
- Identify the audience. Would its readers have a reason to use your resource?
- Check the context. Does the link support a finding, explanation or recommendation?
- Understand the terms. Is the coverage editorial, sponsored or user-generated?
Look beyond one metric before committing money or time. Check the publisher’s recent work, whether the proposed page fits it, the placement terms and the audience you hope to reach. For an existing link, examine the referral visits and enquiries it produces where you can measure them. A placement can be useful to the business even when you cannot demonstrate a ranking change.
Give another publisher something worth referencing
A publisher needs a reason to send readers to your website. Start with something your business can contribute: original data, a case study, a useful calculator or an explanation that is difficult to find elsewhere.
Make the work easy to assess and cite. Explain where the data came from, identify the author and dates, and keep qualifications beside the findings. A small customer sample can still be useful if readers understand its limits.
Before creating another page, look at what you already have. What would help someone researching this subject? My guide to earning natural backlinks without outreach explores how to develop that into a resource other writers can use.
Help the right people discover the work
Contact people already covering the subject. Explain which finding helps their readers and link to the evidence. If a publication already mentions your research or tool without linking, ask whether it could add the original source. The writer decides whether it belongs.
For paid placements, Google recommends rel="sponsored"; nofollow is also acceptable. These links can still bring relevant visitors, but should not be sold as guaranteed ranking credit.
Be wary of packages promising a fixed number of high-score links. Ask where they will appear and why those publications would include them. Google’s link-spam policy covers paid links intended to manipulate rankings, excessive exchanges and automated link creation.
Tools for monitoring backlinks
Search Console is a useful starting point for your own website. A dedicated backlink tool adds competitor research and more ways to investigate new or lost links.
| Tool | Useful for | Keep in mind |
|---|---|---|
| Google Search Console | A free overview of which sites link to your verified website and which pages attract links. | The Links report is a sample, not a complete or live inventory. It does not identify nofollow links. |
| Ahrefs | Investigating new and lost backlinks, referring pages and reasons a link was marked lost. | “Lost” can reflect crawl problems or redirects. Check important placements on the source page. |
| Semrush | Tracking new and lost links, comparing competitors and finding gaps in referring domains. | Report access and limits depend on the subscription. Compare the same scope and filters over time. |
| Majestic | Examining link context, topical relevance and backlink history through its Fresh and Historic indexes. | Historic records may include links that are no longer live. Its authority metrics are third-party scores. |
Each tool discovers and updates links differently, so totals will vary. Use one consistent source to track changes and inspect valuable links directly.
Assess the work beyond the backlink count
Review which publications linked, the pages they chose and whether those links brought relevant visitors or enquiries. Check organic visibility too, while accounting for other changes such as a rewrite or better internal links.
For AI search, use a consistent set of questions and record the platform and date. Track source citations separately from brand mentions. One favourable answer is too little evidence to judge a campaign.
A warning score alone is not a reason to disavow links. Google says most sites do not need its disavow tool; its guidance concerns substantial artificial links alongside a manual action or the likelihood of one.
Plan content that others can reference
I help businesses research useful topics, develop evidence-led content and connect it with their SEO priorities. If you have expertise or data that deserves a wider audience, we can discuss how to turn it into a useful resource.
Discuss your content project