Aydın Nasuh Every few weeks someone on X says backlinks are dead because AI answers don’t need them. Every few weeks someone else sells a link package because “AI citations run on authority”.
DataForSEO’s Search Insider 2026 report tested it with data. It looked at about 1.1 billion backlink URLs, 137.9 million Google AI Overview answers and 27.5 million ChatGPT answers.
The short answer: backlinks still go together with AI visibility. Sites with more referring domains get cited more. But I also have a case that breaks the simple version of that story. A 7-year-old SaaS with 1,812 referring domains, more than most sites AI cites, and AI still barely names it.
When you run a backlink analysis on your own site, look at which pages link to you and how they describe your product.
Here’s what the data says, what the case adds, and what I’d do with both as a small SaaS.
The short version
- Links still matter
- More referring domains, more AI citations. In Google AI Overviews and in ChatGPT.
- Domains beat UGC links
- Total referring domains predict citations better than links from Reddit-style sites.
- Harder topics need more
- Sites cited for the hardest keywords have about 5.6 times more referring domains than sites cited for the easiest.
- It's not cause and effect
- The report says it clearly. Links and citations rise together. Nobody proved one causes the other.
- Count isn't enough
- A SaaS with 1,812 referring domains is named in 1 of 16 AI answers. Its links don't talk about its category.
What 1.1 billion backlinks show
DataForSEO compared each domain’s backlinks with how often AI cites it. Two kinds of links: all referring domains, and UGC links (links marked rel="ugc", usually from forums, comments and community sites).
How strongly each link signal goes with AI citations
All four are moderate and positive. Not weak, not strong. More links usually means more citations, with plenty of exceptions.
The UGC numbers look dramatic on their own. Among sites with more than 10,000 UGC links, 37.6% have more than 500 AI Overview citations. Among sites with no UGC links, it’s 0.19%.
But that doesn’t mean you should go collect forum links. Total referring domains show a stronger link than UGC links, in both engines. The report’s own advice is the same: build a broad, relevant link profile and don’t chase UGC links for the numbers.
Rankings still follow links, a bit
The report also looked at plain Google rankings.
Referring domains of sites in Google's top 10
- Position #1 11168–240
- Position #2 9063–171
- Position #3 8562–161
- Positions 4–10 8561–158
Sites at #1 have a median of 111 referring domains. Sites at #4–10 have 85. That’s 31% more. But look at the ranges. They overlap a lot. Plenty of sites rank #1 with fewer links than sites at #7.
So links help you rank. They don’t decide it alone.
Harder topics, bigger link profiles
This is the part I found most useful. The report checked the referring domains of sites cited in AI Overviews, grouped by how hard the keyword is.
Referring domains of sites cited in AI Overviews, by keyword difficulty
Log scale
- KD 1–20 282236–447
- KD 21–40 371256–732
- KD 41–60 537296–1,326
- KD 61–80 942399–2,619
- KD 81–100 1,571518–5,765
For the easiest keywords, the typical cited site has 282 referring domains. For the hardest, 1,571. That’s 5.6 times more.
Keep that last number in mind: 1,571.
The case: 1,812 referring domains, 1 mention
I’ve written about this tool before. It’s an uptime monitoring SaaS, seven years old, with thousands of users. Its owner let me analyse the data. I’m not naming it.
The tool, October 2026
- Age
- 7 years
- Referring domains
- 1,812
- Backlinks
- About 105,000
- Backlinks from footers
- About 98,700
- AI mention rate
- 1 of 16 answers
1,812 referring domains. That’s more than the median for the hardest keyword group in the report. By the numbers alone, this tool should be cited everywhere.
It’s named in one AI answer out of sixteen. That one is a long list in Claude. In every “best”, “alternatives” and “small teams” question, in every engine, it’s missing. The full numbers are in my LLM visibility post.
So where did the links go wrong? Look at the fourth row. About 98,700 of its 105,000 backlinks sit in a footer. They come from status pages the tool hosts for its own customers. Every customer page links back. That builds a big count fast.
But those links don’t say anything. They don’t call the tool “a good uptime monitor”. They don’t appear in a “best uptime monitoring tools” article. They don’t appear where buyers compare options. They’re a signature at the bottom of a page.
The few links that do talk about its category are old and on small sites. Meanwhile its competitors list each other in their “best tools” articles, and AI repeats those lists.
That’s how I read the report’s numbers now. Referring domains go together with citations because, for most sites, more links means more people writing about you. When the links come from somewhere else, like your own customers’ footers, the count goes up and the talk doesn’t.
What this means for a small SaaS
The data is DataForSEO’s. The plan below is my view.
My own starting point: ShubHQ had 11 external links after four months. Even the easiest group in the report has a median of 282 referring domains. I’m not going to close that gap by counting. So I’m choosing which links to go after.
Go after links that describe you
One mention in a "best tools for X" article beats a hundred links that don't say what you do. These are also the pages AI reads when it builds a list.
Start with the sources AI already cites
Ask your buyer questions in Google and ChatGPT. Note the sites they cite. Those are your first outreach targets.
Use directories as a base, not a strategy
Relevant SaaS directories give you your first referring domains and a consistent description. I'm on 24. They didn't get me into a single AI answer on their own.
Join conversations, don't farm them
Reddit and community threads matter for visibility. But the data says UGC links alone are a weaker signal. Be useful there. Don't treat threads as link slots.
Pick fights you can win
Compare your referring domains with the sites AI cites for your question. If they all have ten times more, start with a narrower question.
Two things I’d skip:
- Buying links by Domain Rating. The report shows correlation, not cause. And the case above has plenty of domains. It doesn’t have the right kind of mention.
- Footer and widget links as an AI strategy. They can help Google find you. They don’t tell anyone, human or AI, what you’re good at.
Quick check for your link profile
- At least one "best tools" article in your category mentions you.
- Most of your links come from pages about your category, not about something else.
- You know which sites AI cites for your main buyer question.
- Your product is described the same way on every site that links to you.
- Footer or widget links aren't most of your backlinks.
FAQ
Do backlinks help you get cited by ChatGPT?
They go together. In DataForSEO's 2026 study, referring domains correlated with ChatGPT citations at Spearman ρ 0.38. The study doesn't prove that links cause citations.
Do backlinks help with Google AI Overviews?
The link is slightly stronger there, at ρ 0.41 for referring domains. Sites cited for the hardest keywords had a median of 1,571 referring domains, against 282 for the easiest.
Are links from Reddit and forums better for AI visibility?
Not on their own. UGC links correlated with AI citations, but total referring domains correlated more strongly in both Google AI Overviews and ChatGPT.
How many referring domains do I need to be cited in AI Overviews?
There's no fixed number. In the study, the middle half of cited sites ranged from 236 to 447 referring domains for the easiest keywords and from 518 to 5,765 for the hardest. Some sites with far more are never cited.
Why does a site with many backlinks still not appear in AI answers?
Count isn't context. If most links don't describe your product or category, like footer links from customer pages, they add little to how AI sees you. Mentions in category articles and comparisons matter more.
Where I am
ShubHQ, November 2026
- External links (first 4 months)
- 11
- SaaS directories
- 24 so far
- Mentions in other people's category articles
- Very few. This is the next job.
- LLM visibility
- 0 of 16 AI answers name ShubHQ, tracked since Sep 27
I'll update this box as category mentions come in, and we'll see if the AI number moves with them.
See which sites AI cites for your category. ShubHQ asks your buyer questions in ChatGPT, Gemini, Grok and Google AI Overviews, saves every answer and lists the sites it cited. Those sites are your outreach list. Create a free account → Or run the free LLM visibility checker first, no signup.
Sources
- DataForSEO: Search Insider 2026 Edition (October 2026). ~1.14B backlink URLs, 137.9M Google AI Overview answers, 27.5M ChatGPT answers.
- ShubHQ AI Visibility, anonymous 7-year-old uptime monitoring tool, category “Uptime Monitoring Tools”, Sep 16 – Nov 1, 2026
- ShubHQ Google Search Console and backlink data, own site, first four months
