Original measurement
One of the companies an AI assistant named has a single referring domain. Off-page authority is not the gate.
We took 9 companies that ChatGPT actually named in our runs and measured their off-page strength. The range runs from 1 referring domain to 2,095 - a spread of more than two thousand times, inside a set the assistant itself selected.
1
referring domain on the weakest company an assistant named, against 2,095 on the strongest
Key findings
- Among 9 companies an AI assistant named, off-page strength ranged from 1 referring domain to 2,095 - a spread of more than 2,000 times.
- The weakest company in the named set had 1 referring domain, 1 backlink and 20 organic keywords, and was still named.
- Authority cannot be a requirement for being named by an assistant if a company with 1 referring domain is in the named set.
- Of 9 companies an assistant named, 5 also ranked in Google's top 20 for a keyword close to the question they were named on - a partial overlap, not a rule.
How this was measured
- When
- Off-page figures pulled 31 August 2026; the naming runs they refer to were made 30 August 2026.
- Instrument
- Third-party estimates of referring domains, backlinks and organic keyword counts, plus a ranking check for a keyword close to each company's naming question.
- Surface
- The companies are those ChatGPT named in our 15 service-intent questions. This is a set the channel chose, not one we assembled.
- Sample
- 9 companies - every named company that resolved to a readable domain.
- Pre-registration
- The falsification test and its thresholds were written down before any figure was pulled.
The test, and why it needs no control group
A fair control group of comparable companies that were not named does not exist in a form anyone can trust. So the test was built as a falsification on the minimum, which needs no control at all:
Authority cannot be required for being named if even one named company lacks it.
The threshold was fixed in advance: if the weakest named company had fewer than 50 referring domains and fewer than 100 organic keywords, off-page authority is not a requirement. If all 9 had more than 500 referring domains and more than 1,000 organic keywords, it plausibly is one.
What we found
| Company | Referring domains | Backlinks | Organic keywords |
|---|---|---|---|
| A | 1 | 1 | 20 |
| B | 51 | 266 | 128 |
| C | 79 | 437 | 35 |
| D | 104 | 418 | 95 |
| E | 115 | 614 | 28 |
| F | 178 | 1,185 | 15 |
| G | 451 | 5,890 | 287 |
| H | 466 | 7,942 | 258 |
| I | 2,095 | 29,935 | 2,000 |
Nine companies, ordered by off-page strength and shown without names: these are third-party estimates rather than our own measurement, and we do not attach an estimate to a named company. The naming observations themselves are our own and are named openly elsewhere on this site. Median: 115 referring domains.
The verdict the thresholds produce
Company A sits at 1 referring domain, 1 backlink and 20 organic keywords - comfortably under both pre-registered lines. On the test as written, off-page authority is not required to be named.
That is a narrow claim and it is worth stating precisely. It does not say links are worthless, that authority never helps, or that a stronger domain is not named more often. It says that a company with essentially no off-page footprint was in the answer, so the absence of one is not what keeps a company out.
It also decided where money goes. A link-building and PR budget from day one cannot be justified on the grounds that assistants require authority, because the set the assistant picked contains a company that has none.
The second test: does naming follow Google ranking?
For each company we also checked whether it ranked in Google's top 20 for a keyword close to the question it was named on. 5 of the 9 did; 4 had no close ranking at all, and 1 sat at position 48.
The pre-registered reading for 5 of 9 is explicit: partial - report per company and do not generalise. 6 or more would have meant naming follows ranking. 2 or fewer would have meant it is independent. 5 is neither, and we are not going to round it into a story.
What can be said is that ranking is clearly not necessary: 4 named companies had no nearby ranking whatsoever.
What this means if you are buying
- If you have been told that being recommended by AI assistants requires a strong backlink profile first, that is not what the named set looks like.
- For a small business or a small agency, this is the encouraging finding in our data: the entry ticket to this surface is not a budget you probably do not have.
- It also narrows what is left. 4 other page-side explanations failed the same way, which is why we now work on owning the specific facts an answer needs rather than on the shape or strength of the site around them.
Honest limits of this measurement
- 9 companies is a small set, and two of them were named for the same trade, so the sample is not fully independent.
- The off-page figures are third-party estimates, not ground truth. They are good enough to separate “one referring domain” from “two thousand”, which is all this test needs, and not good enough to rank companies against each other.
- One point in time. A company's authority on the day we measured is not necessarily its authority when it was named.
- A falsification on the minimum answers whether authority is required. It does not answer whether authority helps, and we make no claim either way.
Questions about this research
Are you saying backlinks are useless?
No. We are saying they are not the gate. A company with 1 referring domain was named, so the lack of a link profile is not what keeps a company out of an assistant's answer. Whether links improve the odds is a separate question our data does not settle.
Why are the companies not named in this table?
Because these are third-party estimates rather than our own measurement, and attaching an outside estimate to a named company would be presenting someone else's approximation as a fact about that business. The observations we made ourselves - which companies an assistant named, and when - are published openly.
Does this apply to Google search too?
No, and we do not claim it does. This is a measurement of which companies an AI assistant names. We do not measure or diagnose classic Google rankings, and nothing here should be read as advice about them.
Then what does decide who gets named?
On our evidence, owning the specific fact an answer needs. 5 separate page-side and off-page explanations failed to predict naming; what the cited pages had in common was that each held the concrete information the question required.
More measurements from the same instrument
- Five things we tested that do not predict whether an AI assistant names you
- An AI answer to a business owner cites 8.7 sources
- All seven measurements →
We run this measurement for a living. If you want it run on your own category and your own city, the free scan is the same instrument on one question, and the report is the full set. Nothing here is behind a form.
Every category we measure has its own page, with the questions customers put to an assistant in that category and what a reading costs. See all industries we measure →
Published 31 August 2026. Every figure on this page is recomputed from raw observation files by a single script, so any of them can be traced back to the readings behind it. If you are named here and believe a reading is wrong, tell us: we will run the measurement again and publish what it returns, whichever way it goes.