How the Trust Score is calculated
Algorithm v1.3.1
We built this to be checkable, not just a number. Every score you see comes from the exact method below — the same categories, the same weights, the same verification steps, run against the site you entered. Nothing is hidden or hand-adjusted per site.
Social proof & client claims
25% of scoreLogos in a trust bar, named clients, or 'we've worked with X' claims, checked for a real, findable relationship.
- Every logo or named client is treated as a claim: "this company is a client or partner."
- We check whether the site links the claim out to independent evidence: the logo or name linked to the client's own website. Where web search is available, we also look for an independent public page connecting the client to the site.
- Both signals together score highest (verified). Either one alone scores well (likely true). A claim with neither scores low (unverified): it may still be true, it just isn't publicly checkable.
Team & company facts
20% of scoreClaims like team size, years in business, or founding date, checked against the domain's registration age and the company's public LinkedIn page.
- Team-size claims ("a 200-person team"): where web search is available, we look for the company's public LinkedIn page so the headcount can be compared. Without one, a team-size claim stays unverified.
- Founding-year and years-in-business claims are checked against the site's actual domain registration age (from public RDAP records). A claim within about 1.5 years of the domain's age is verified; one that exceeds it by more than 2 years is flagged as contradicted. A person's own years of experience is not treated as the company's age.
- Office and location claims are compared with any address shown elsewhere on the site. A match confirms the site is consistent about its address; we don't look the address up on a map.
Testimonials & reviews
20% of scoreQuoted testimonials and review scores, checked for attribution (a real, findable person) rather than anonymous or stock quotes.
- Each testimonial is scored on attribution: a full name, role and company scores well (likely true); a named person whose LinkedIn profile is linked from the page scores highest (verified); first names, initials or anonymous quotes score lowest (unverified).
- Aggregate review scores (Google, Clutch, G2 and others) are checked for traceability: does the site link to its profile on that platform, or, where web search is available, does a public profile turn up? We don't read the platform's own rating, so compare the two numbers yourself.
Credentials, certifications & awards
15% of scoreBadges, certifications, partner-tier claims, and awards, checked for a link to the issuing body or the award announcement.
- Certifications, partner badges and awards are checked for whether they link out to the issuing organisation or the award announcement, rather than sitting on the page as an image (likely true).
- Where web search is available and nothing is linked, we confirm the programme or award exists publicly, but confirming this company's membership still needs a manual check of the issuer's directory, so the claim stays unverified.
Transparency & contactability
10% of scoreWhether the site is straightforward about who runs it: a real address, working contact details, and named leadership.
- Every site gets a baseline read from what's actually on the page: HTTPS, a linked privacy policy, a linked terms page, and a visible phone number, email address, and mailing address — this alone accounts for most of the category even when the site makes no explicit claim about itself.
- Where the site does make an explicit claim (naming its leadership, or stating its contact details), that claim is verified on top of the baseline: named leadership is checked for a linked, findable profile (e.g. LinkedIn); an unnamed 'our team' claim scores lower than a named one.
Internal consistency
10% of scoreWhether the numbers and claims the site makes agree with each other from page to page.
- Figures that measure the same thing (team size, years in business, clients served, projects delivered and so on) are compared across every page we read. Figures are matched by what they measure, never just because they're numbers, and percentages (a share of some group) aren't compared.
- Exact figures must agree within 10% (allowing for rounding); a figure like "500+" counts as a lower bound, so it only conflicts with an exact figure clearly below it. A contradiction lowers this category's score and is listed as a flagged discrepancy.
How the categories combine
Each claim is checked and given one of four statuses: Verified (confirmed against an authoritative source), Likely true (a credible independent signal exists, short of full confirmation), Unverified (no public signal either way — not proof of anything false), or Contradicted (public information conflicts with the claim). A category's score is the average confidence across its claims, with two exceptions: Transparency blends its claims (60%) with what's visible on the page (40%), and Internal consistency is the share of repeated figures that agree with each other. If a site makes no claims of a given type — say, no awards section — that category is left out of the average entirely rather than scored as zero, so the overall score reflects what a site chose to claim, not what it chose not to.
What this isn't
This is an automated read of public information, not a legal or financial audit, and it doesn't access private systems (LinkedIn/Glassdoor logins, internal HR data, and similar) to confirm anything — only what's publicly searchable. An "Unverified" claim may well be true; it simply isn't independently checkable from the outside, which is itself useful to know if you're the one making the claim.