I've been building Linkbaits.com for the past year and a half. The core of it is a database of 204 pages that each have 10,000+ referring domains. Here is what the data actually says — because most of what's written about link building is theory. This is data.
We tracked 204 pages that each have 10,000+ referring domains. Format predicts performance better than quality, embed codes multiply links 3–5x, and the citable unit is the critical variable.
According to Victorious, a Q3 study of AI Overviews found that only 37% of the pages AI Overviews cite rank in Google's top 10 — while AI Overview appearances jumped from 72.8% to 97.5% in one quarter. Read the full study: AI Overviews and organic rankings.
That finding is the reason this dataset matters more now than when the tracking started. The pages AI systems cite are not simply the pages that rank first — they are the pages with the structural properties in this article: a citable unit, an embed layer, a format that gives writers a reason to link. The 204 pages below are, in effect, the training set for what gets cited in the AI-answer era. If you want your pages in the next version of that set, build what these pages built.
The database covers 204 pages, each with more than 10,000 referring domains. Referring-domain counts were pulled from backlink-data sources including DataForSEO. Each page was classified by format, checked for an embed layer, checked for a citable unit, and tracked for early link velocity. The four findings below held across formats, niches, and page ages.
The median referring domains by content format:
67x gap between free tools and blog posts. These are not low-quality blog posts — I specifically selected pages that were professionally produced. The gap is structural, not qualitative.
The highest-performing formats all share one property: they give writers a reason to link that is independent of content quality. Tools get linked because writers say "use this." Research gets cited because writers need a primary source. Guides get linked because writers prefer to reference rather than re-explain.
Blog posts rarely have these structural link incentives. They are designed for readers, not for writers. The cost side tells the same story: a free tool costs roughly $8.80 per link, original research roughly $13.80, digital PR roughly $44, and a guest post roughly $176. Format choice is a budget decision disguised as a content decision.
Full data: the Linkbaits 2026 guide.
This was the finding that surprised me most.
Of 204 high-performing pages, 63 had copy-paste embed codes — either for charts or for embedded tools. Those 63 pages earned disproportionately more referring domains than comparable pages in the same format without embed codes.
The mechanism is obvious in hindsight: a writer who embeds your chart or tool directly in their article creates a referring domain without you doing any outreach. The embed code is a passive link-earning machine.
And only 31% of even the top-performing pages had them. In the broader web, the percentage is far lower.
If you are building any data visualization, free tool, or interactive feature and not adding an embed layer — you are leaving 3–5x of your potential referring domain count on the table. This is the cheapest finding in the dataset to act on: an embed code is an afternoon of work. Check any page you own with the free backlink checker before and after adding one, and the effect is usually visible within two months.
The pages that earned the most links in months 1–3 earned the most lifetime links. The correlation was strong enough that early velocity was a better predictor of 24-month link performance than format type.
The reason: initial links trigger ranking movement. Rankings drive organic discovery. Discovery drives organic links. Organic links improve rankings further. The compounding loop only starts if there is enough initial velocity to trigger the first ranking movement.
Pages that published and waited — no journalist pitches, no community posts, no email list announcement — earned approximately 10% of the lifetime links earned by pages with active month-one distribution.
This means the return on seeded distribution is enormous relative to its cost. 20 journalist pitches and an email list announcement cost almost nothing compared to the additional organic link velocity they unlock. It also means "publish and pray" is not a strategy with a low success rate — it is a strategy with a measured 90% penalty. The citation chain explains why: without the seed, phase two never starts.
Across all 204 pages, the single most reliable predictor of link volume was whether the page had a citable unit — a specific, quotable finding that another writer could drop into their article and attribute to a named source.
Pages with a clear citable unit earned a median of 3.2x more referring domains than equivalent pages in the same format without one.
This held even for free tools: tools with companion data articles that included a citable finding (for example, "across 10,000 users of this tool, the median result is X") earned more links than identical tools without companion data.
The citable unit is not a writing decision. It is a research design decision made before data is collected. You have to decide what specific number you want the research to produce, then collect the data to produce it credibly. Before your next build, run the concept through the 30-second test: if you cannot name the structural reason a writer would link to the URL, you do not have a citable unit yet.
Four findings, one build order. Pick the format with the best economics you can afford — tool first, research second, guide third. Design the citable unit before you collect a single data point. Ship the embed layer on launch day, not as a later addition. Seed with 20 journalist pitches, one email announcement, and three community posts in week one. Then measure monthly: referring domains, new domains per month, and whether your commercial pages are moving — the authority halo is the payoff that justifies the whole program.
All 204 pages, the full format data, and the 8-element structural checklist are published at the Linkbaits 2026 guide.
The AI tool at linkbaits.com identifies citation gaps in your specific niche and generates 10 linkbait concepts designed around these findings — with format recommendation, expected link range, and a brief structured around citable unit design. Browse the AI link-building tools or the linkbait database.
Free at linkbaits.com. Happy to answer methodology questions in the comments.
Get 10 concepts built on these findings
Enter your niche and Linkbaits.com finds citation gaps, recommends the right format, and drafts your citable unit — in 30 seconds.
Start free →