Methodology

The Public Evidence Score rates link building providers from observable public evidence alone on stated dates. Every input is checkable by a third party from public sources on a stated access date. The full specification is versioned as PES v1.2 (2026-09-07), and every article on linkbuilding.wiki is built against that version.

Criteria and weights

Eight criteria feed the score, each carrying a fixed weight, and the eight weights sum to one hundred, as listed below.

• E1, Corporate identity and longevity, weight 10.
• E2, Price transparency, weight 12.
• E3, Published terms and buyer protection, weight 12.
• E4, Independent reputation, weight 18.
• E5, Verifiable delivery evidence, weight 13.
• E6, Claim integrity and disclosure, weight 10.
• E7, Operational footprint, weight 5.
• E8, Value for money across a price basket, weight 20.

Rubric bands

Each observed criterion takes an integer band from zero to four, with four as the top. Graders award the best-supported band against a written rubric, because bands are grades and two graders should reach the same result.

• E1 rewards a registered entity matched to the trading name, with named leadership and a domain first archived several years ago.
• E2 rewards transactable prices with currency and stated inclusions for each tier; quote-only selling and missing prices sit low.
• E3 rewards public terms covering refunds, replacement cover, turnaround, cancellation, and a dispute route; missing terms sit low.
• E4 combines review ratings across platforms, weighted by review count and smoothed toward the round mean, with recency and flags recorded.
• E5 rewards live sample placements on third-party domains plus checkable case studies and a stated sourcing model; screenshots alone sit low.
• E6 starts at band four and falls one band per adverse finding, to a floor of zero. Adverse findings are guaranteed-outcome claims, a contradiction between the provider's own pages, and undisclosed paid placement. Absent policy statements on link attribute, sponsored disclosure, or private-network stance cost at most one band in total.
• E7 rewards headcount, a named team, several contact routes, turnaround statements, and a physical office; bare contact forms sit low.
• E8 grades each basket item with a published price against the cohort median adjusted price. Item grades then average to one decimal place for the provider. A published price for a defined unit states the amount and currency with a unit definition that names the authority floor. Transactability sits under E2 alone, so published prices grade under E8 whether or not the order is transactable. Quote-only providers with no published unit price remain not observed on E8.

The silence rule

Where silence is itself the evidence, the criterion stays observed and earns a low band. Missing prices, terms, delivery evidence, or disclosures therefore sit at the bottom of their scales. Where silence reflects the limits of public observation, the criterion is marked not observed and excluded from both parts of the calculation.

• Always observed: E2, E3, E5, E6. No price means band zero for E2, and no terms means band zero for E3.
• Not observed when uncheckable: E1, E4, E7, E8. E4 needs at least five reviews in total; E8 needs a published price for a defined basket-item unit.

Nothing is filled with a neutral value, so the published coverage states exactly how much of the method rested on observed evidence.

Evidence coverage and the 70 threshold

Observed weight sums the weights of observed criteria, and the raw score scales the band-weighted sum to one hundred. Coverage equals the observed weight on a scale from zero to one hundred, before penalties subtract after rescaling. Penalties subtract after rescaling, capped at thirty points in total, and the result rounds to one decimal for display.

A provider is ranked only when its coverage reaches seventy or more in the round. Below that line the provider appears under "Listed, not ranked", with the missing criteria named. Such providers carry no rank number, and publishing the missing evidence is the route into the ranking.

The price basket and deliverable value index

E8 compares like with like through a frozen reference basket fixed for the round. The basket holds four items, with definitions fixed for the round in the list below:

• G30, a guest post at authority floor thirty with provider-written content.
• G50, the same placement at authority floor fifty.
• N30, a niche edit at floor thirty in an existing article.
• M1, the cheapest monthly package divided by its minimum links.

Each basket item assumes provider-written content and one dofollow link unless the listing states otherwise. Listed prices convert to dollars at the round's frozen exchange table before adjustment for comparison. Each price is then adjusted by a deliverable value index that starts at unity and multiplies per attribute. Stated traffic floors and higher authority floors raise the index, while longer content and longer cover raise it slightly. Nofollow or sponsored status halves the index, and a written topic-match promise raises it slightly.¹ The adjusted price for comparison equals the dollar price divided by the deliverable index.

Each item is graded against the cohort median of adjusted prices for that item, with bands assigned by ratio as follows:

• Ratio up to 0.70 takes band four; up to 0.90 takes band three.
• Ratio up to 1.10 takes band two; up to 1.40 takes band one.
• Ratio above 1.40 takes band zero.

The E8 band averages those item grades to one decimal place for the provider. The specification's worked example prices G30 at ninety dollars with content included and no traffic floor. Six-month replacement cover applies, giving an index of unity and an adjusted price of ninety dollars. A second seller prices the same item at one hundred eighty dollars with a five-hundred traffic floor. Otherwise identical terms give an index of 1.15 and an adjusted price near one hundred fifty-seven dollars. Against a median of one hundred twenty, the first ratio near 0.75 takes band three and the second near 1.30 takes one.

Penalties P1–P6

Each penalty below sits outside the bands and subtracts from the rescaled score instead. Each one is evidence-gated, listed by name on the article, and capped at thirty points in total. Status values are applied, pass, or unknown, where unknown means the evidence could not be read at all.

• P1, No placement pre-approval, 15 points. It applies when no public route lets the buyer approve the placement domain before publication. A route offered as standard or on request passes, and quote-only sellers pass only when a client approval step is described publicly.
• P2, Guaranteed outcome claims, 10 points. It applies when public copy guarantees rankings, traffic, or authority-metric increases.²
• P3, Same-surface contradiction, 8 points. It applies when a headline promise is contradicted by footer, legal, or terms text on the same site.
• P4, Owned-network placements sold as editorial, 20 points. It needs evidence such as shared registrants, hosting, templates, or the provider's own statement. Suspicion alone stays in Talk as alleged.
• P5, Review-platform integrity flag, 8 points. It applies on any consumer alert, incentivised-review flag, or platform sanction found on a profile.
• P6, Undisclosed non-dofollow default, 10 points. It applies when a sold backlink defaults to nofollow or sponsored without saying so.¹

Only P1 treats total silence across pages, terms, and order flow as an applied penalty. The article then states that no such route was found as of the access date.

Category evidence and category rank score

Each service and industry category scores four items on the same zero-to-four bands with equal weight, defined as follows:

• C1, a dedicated public offer page for the category.
• C2, category-specific published price or scope.
• C3, category-specific samples or case studies.
• C4, category-specific terms, restrictions, or publisher rules.

The Category Evidence Score scales those four bands to one hundred for the category. A provider is eligible only with a genuine category page and a ranked overall score. Genuine means the first item reaches at least band two on the four-band scale. The category rank score blends seven tenths of the Public Evidence Score with three tenths of the Category Evidence Score.

Two service categories describe the whole trade rather than a distinct product: link building services and link building packages. For these umbrella categories, a provider's Category Evidence Score takes the higher of two values. The two values are the score graded for the umbrella category itself and the strongest score earned in a donor category. The donor categories are the link placement and link earning services: guest posting, niche edits, high authority links, white label link building, digital PR, journalist requests, and local link building. Press release distribution, AI visibility, and local citation building are not donors. A provider is eligible for an umbrella table through the umbrella category itself or through any donor category, and each umbrella row records which category supplied its score. Industry categories keep their existing treatment, with no umbrella adjustment applied to their tables.

Ties

Providers order by category rank score, with the highest score taking the first position. Providers within one point of the leader share the rank, ordered alphabetically within the tie. Later ranks skip accordingly, so a two-way tie for first is followed by rank three. Ordering beneath the display uses full precision rather than the rounded public figures shown.

Rounds and rescoring

Scoring runs in rounds, each carrying an identifier that records the month and sequence. Evidence and scorecards live in append-only round directories, and a rescore is always a new round rather than an edit. The round file stores the frozen exchange table, the reputation prior mean, and the basket medians used for comparison. Version 1.2 (2026-09-07) replaced version 1.0 before the first round was published, so no published score has changed.

PES v1.2 (2026-09-07) added the umbrella-category rule for general services and packages tables only.

History entries

History is written by build comparison between data files, never by hand-typing changed figures. Each build compares newly computed results against the last published data file in full. Every changed band, penalty, score, or rank gains its own dated history entry automatically. Manual editorial entries from the revision log stand alongside the machine entries, each naming its editor and reason. Provider articles therefore show the full movement of every published figure across all rounds.

See also

• Editorial policy
• Link building
• Guest posting