Yes, but not for the reason link sellers give. In AI search a link does two jobs, and neither is "passing authority": it gets your page discovered by the crawler, and it makes your claim appear in a place that is not yours. The second job is the valuable one, and it is why a hundred links from the same network are worth less than three from unrelated sites.
Discovery: the unglamorous half
A page nothing links to is a page a crawler may never reach, and a page that was never crawled cannot enter the candidate set for any answer. This is mechanical, not a ranking factor — it is the difference between existing and not existing as far as retrieval is concerned.
Corroboration: the half that decides citations
A claim on one site is a claim. The same claim on eight unrelated sites is treated as settled. That is what the selection step rewards, and it has one consequence that reorganises everything:
Independence counts, volume does not. Eight sites owned by the same person, running the same template, cross-linking to each other, are read as one source repeated eight times. Three genuinely separate sites outweigh them, because they can disagree with each other and did not.
What is worth what
| Kind of link | Helps discovery? | Helps corroboration? |
|---|---|---|
| Mention in an independent article | Yes | Yes — the strongest thing available |
| Listing in a real directory your buyers use | Yes | Some; it confirms you exist and what you do |
| Forum or community answer | Yes | Yes, where it is first-hand and honest |
| Links between pages you own | Yes, usefully | No — a site cannot corroborate itself |
| Bought links from a link farm | Barely | No, and the pattern is detectable |
| Comment spam | No | No |
The row people get wrong
The fourth one. Internal links are genuinely useful — they are how a cluster of articles reads as a cluster, and how a crawler finds your deeper pages. What they cannot do is corroborate, because the source and the subject are the same party. A site vouching for itself adds no information, which is exactly the test the selection step applies.
What this means in practice
The useful target is not a number of links. It is: how many independent domains currently say what you do, in words a retriever can lift? If the answer is zero, that is the problem to solve, and no amount of on-page work substitutes for it.
This is the specific job Magneth does: it publishes your site across hundreds of partner sites with structured links, so references exist on domains that are not yours. Stated plainly, because it is the honest limit: those are partner sites, so they corroborate that you exist and what you offer — they are not a substitute for someone independently reviewing you, and no service can sell you that.
Three things not to do
- Buy a thousand links cheaply. The pattern is detectable, the sites are related, and the whole batch reads as one source.
- Chase a single prestigious link for months. Three ordinary independent mentions beat one trophy, because corroboration counts sources, not prestige.
- Treat link count as the metric. Count distinct domains that describe what you do. That number is the one that moves citations.
