Methodology v1.2 · eligibility v1.3
The public Web Score is independently assessed by LocalCite from public evidence. If we did not open a page, that fact stays unknown. Blocked is not absent. This page is the scoring record, not a sales page.
Primary measure. From opened independent review hosts only. Strongest opened rating (scaled to 100, up to 55), volume slice from the stated review count (up to 30), recency if a review date is within about 18 months (10, or 2 if stale). Opened complaint hosts with a negative public signal can subtract up to 20. Homepage “50 Google reviews” self-claims are not a Maps count. If no opened review page states a rating or count, Feedback is unknown and overall stays blank.
Can a stranger find the current brand on the hosts this local market actually uses? Points from an opened official site, opened Maps (including Maps via DataForSEO), opened directories, extra profession hosts, Instagram/LinkedIn, and a named organic SERP or AI Overview. A blocked host is not scored as missing and is not scored as present.
One name, one NAP-like cluster across opened present/stale pages. A leftover address while the site moved is a hole. Too many distinct display names, or stale/ghost listings, pull the score down. Not enough opened pages leaves this unknown.
Scored only from an opened official page: live site, city and profession in the title, name/city/profession on the page, LocalBusiness/Person schema, a visible contact path. A homepage Google-review self-claim costs 5. A blocked official site is unknown, not a dead site.
Named on a press page or listicle we opened. A thin vendor profile is not “they made the Top 10.” If press queries ran and named nothing, the score is 0, not invented praise.
Known dimension scores are weighted and rounded. If Customer Feedback is missing, overall is null. The public overall never includes private files.
SERP snippets are not opened-page facts. Maps, Yelp, and other hosts can be BLOCKED (timeout, bot-wall, rate-limit). Blocked is not scored as missing and is not scored as present. ChatGPT and Perplexity are not queried live. We never invent ratings, counts, quotes, phones, prices, or ranks.
The engine records confidence as opened rows (present, absent, or stale) divided by planned source rows, as a percent. BLOCKED rows are not counted in the opened numerator. Example on a published certificate: Tess Leslie Photography lists Research confidence 78 — 14 of 18 priority sources. We do not invent a confidence figure when one was not printed.
Gold is not expected. Gold is not “best” and not “no work left.”
A Certificate of Local Standing is issued only when all of the following are true:
Otherwise the outcome is Not Certified (score 0–39) or Insufficient Evidence (confidence below 65%, or overall blank). Payment still buys the assessment. The certificate is valid 90 days from the audit date. The two published Austin Gold certificates keep the validity dates already printed on them.