Mass-link buying never produced great SEO. For AI search it is actively destructive — AI extractors treat low-quality outbound graphs as a negative signal. A site with 500 generic guest-post backlinks scores worse for AI citation than a site with 30 hand-picked donors from AI-trusted sources.

That is a different mental model. AI search is not “more links is better”. It is “every link is justified individually, or skip it”.

The six-criterion scoring model

We developed this model on the AP Education engagement and have refined it across 30+ subsequent engagements. Six criteria, each a hard gate:

Theme relevance. Direct topical match, or a related vertical with editorial credibility. A linen brand can take a sustainability-publication backlink. It cannot take a fintech publication backlink without setting off the topical-mismatch flag.

Traffic ≥ 5k organic / month. Below 5k, the donor is not generating enough crawl signal for AI to weight it. Above 5k it counts; above 50k it counts more.

AIO presence. The donor must already surface in Google AI Overviews on at least some queries. This is the single highest-signal criterion because it confirms AI already trusts the source for retrieval.

Outbound : inbound link ratio ≤ 20:1. A donor with 100 outbound for every 1 inbound is a link farm. AI extractors detect the pattern. Healthy editorial sites run at 10:1 or better.

Domain Rating (DR) ≥ 20. Below 20 the donor lacks the backlink graph to confer authority. Above 20 it counts. Above 50 it counts much more.

Trust Ratio (TR) 0–2. Trust Ratio is the ratio of toxic backlinks to clean backlinks. Below 2 means the donor’s own backlink graph is clean enough not to drag down whatever it links to.

Domain age ≥ 3 years. Younger domains carry too much volatility. Three-year minimum is conservative; some niches require five.

A donor must pass all six gates. Five-of-six is not good enough. We have seen the math fail repeatedly when teams compromise on one gate “just this once”.

What it costs in time

Two to four hours per donor to verify all six criteria. That is why the donor pipeline is a quality-not-quantity operation. We typically place 4–8 donors per quarter on a Scale engagement, 8–16 on Enterprise. Above that the criteria start slipping.

The flip side: a donor that passes all six is worth 5–10× the citation lift of a generic guest post.

AI-trusted sources

Beyond editorial backlinks, three categories of AI-trusted source carry disproportionate weight:

  • Wikipedia and Wikidata — direct editing is constrained, but a brand with a clean entity and a Wikidata Q-number gets weighted 10× a regular backlink for citation purposes
  • Niche industry indices — “top 50 X firms 2026” directories that AI systems crawl as ground truth. Placement here is editorial; payment for inclusion violates the gate
  • Tier-1 vertical publications — TechCrunch for tech, FT and Bloomberg for finance, Lexology and JD Supra for legal. Placement requires real expertise and original commentary

Featured.com, sourceofsources, Qwoted (the HARO-style services) are how most of our clients land tier-1 placements. Three to five published quotes per quarter per named expert is the cadence we target.

What we will not do

  • Mass-buy links from any guest-post network
  • Place links on sites with TR > 2
  • Skip the AIO-presence gate to keep volume up
  • Co-author content on a donor without a real editorial relationship

The six-gate model is non-negotiable on every Answerly engagement. Clients sometimes ask for “more links faster”. We say no, then explain why.

What you should do this quarter

If you have an SEO retainer that is doing link-building without articulating gates like these — pause that line item. The links you are buying are most likely net-negative for AI search.

Start instead with the AI-trusted source layer: get the named experts onto Featured.com or Source of Sources, target three to five quoted-source placements per quarter, build the entity authority graph one verified placement at a time.

If you want the scoring model implemented for you: that is what Scale ($4,800 / month) and Enterprise ($8,900 / month) include. Donor selection runs against the model; every placement justified individually; no exceptions.