BV score methodology: how a backlink's value is worked out

Written by Palash Bagchi, founder of bklink, who maintains mxverify.

The BV score is worked out the same way for every site, from two things: how strong the site is and what kind of link it gives. Every figure on mxverify is measured by an audit, read from the site itself, set by a published rule or taken from a source it names. Nothing is supplied by the sites, and no site pays to be listed, valued or left out.

Where the measurements come from

Links enter the record when someone audits them on bklink.uk, in one of two ways. A placed link comes from a list the customer gives — typically an agency or a brand checking what a PR distribution or guest-post vendor delivered. A scan-found link comes from a backlink scan of the customer's own site: every page found linking to it, placed or not. Both are checked the same way: the auditor fetches each page and reads the link itself. Only placed links say how a publisher treats a link it was given, so only they count toward a publisher's record (below). Who ran an audit, and whose site the links point to, is not published — unless that site's owner chooses to publish it as backlinks audited.

Link behaviour

Each audited link is counted once, at its most recent audit, in one of five states:

• Followed — the link is on the page with no nofollow, sponsored or ugc attribute.
• Nofollow or sponsored — the link is present but carries one of those attributes. Search engines treat these attributes as hints; BV counts such a link as passing no ranking credit.
• Removed — the page loads but the link is no longer on it.
• Dead — the page is gone (a 404 or 410, or a page that says it is not found).
• Could not read — the site blocked or timed out our auditor. These are excluded from percentages rather than guessed.

Percentages are shown only when at least 3 audited links on a publication could be read. Below that, the counts are shown and the rates are not.

Shared servers and networks

Each publication's domain is resolved to the address it serves from, and that address to its hosting network and country. Publications that resolve to the same address are listed together, and a server page exists where 10 or more do. CDN edge addresses (Cloudflare, Fastly, Akamai) are never grouped: they front unrelated sites by design.

A shared address alone is not a network. Large hosts put unrelated sites side by side, and one syndication host can carry the finance pages of hundreds of real newspapers. So each listed server is checked for five footprints of one owner, all from public records:

• Nameservers — 60% or more of its sites use the same privately run nameservers. A registrar's or host's defaults (Cloudflare, GoDaddy, Namecheap, Dynadot and the like) are shared by millions of customers and do not count.
• Registration dates — 60% or more of its registered domains were registered inside one 90 days.
• Registrar — one registrar holds 70% or more of them.
• Site theme — one WordPress theme runs on 70% or more of at least five homepages read.
• Audited purchases — 90% or more of its sites carry the same number of audited links, and 95% or more of those links share one follow state.

A server is a network when at least two footprints agree and one of them is nameservers, registration dates or theme — the ones a popular registrar or a buyer's habits do not produce by chance. It is a syndication platform, never a network, when 90% or more of its sites are subdomains of fewer than ten registered domains, or its domains were registered years apart with a median year before 2010. Where only the registrar and the buying pattern agree, it is marked as one owner possible and is not treated as a network. Dates and registrars count once per registered domain, so one domain's many subdomains are one registration. Up to 40 sites per server are read, across as many registered domains as it holds; each site is read once and again after 90 days, and the verdict is recalculated from those readings as audits add sites. BV values a link from a network's site at level D. Each server page shows its footprints and verdict.

Authority

Half of a site's strength in the BV score is its Ahrefs Domain Rating, read through the Ahrefs API and refreshed at most every 30 days. mxverify uses it as an input to the value and does not display it; strength is shown in steps of 5, so the rating cannot be read back from it.

Backlink value

The backlink value checker gives any website a BV score: the dollar value of one new link from it under BV 2.9, the model bklink uses to value every link in an audit (full method). It combines the site's strength — half its Ahrefs Domain Rating, half an estimate of its organic visits — with the type of link it usually gives, read by a model from the site's audited links where there are any, otherwise from our own reading of the site, otherwise set by fixed rules. A search engine, social network, video platform, messaging app or AI assistant is priced at the link it usually gives — a listing, profile or user post — and its report lists every kind of link it can give, with which of them are not backlinks at all, in place of the four levels used for publications. A site you look up is recorded for 30 days, so the next lookup — and any bklink audit that meets it — reuses the figures instead of buying them again. It gets a site report at the same address a publication would have, marked valued, and is counted as a publication only once an audit has measured links on it.

Traffic

Traffic figures are estimates from third-party traffic data, as on every traffic tool; no one outside a site sees its analytics. A site's report charts a third-party estimate of its traffic — visits, channels, countries, keywords and AI-assistant referrals — where one is on record. Sites met in a bklink audit carry another third-party estimate of organic visits. Each report dates the figure it shows, and BV prices on the organic-search figure alone. A site with no figure on record counts as no traffic until one is fetched: the median measured site has about one visit a month, so counting a typical site's figure would mark it up.

What each link level means

Every report prices a new link from the site at one of four levels. The level is the kind of link, and its factor scales the value; nofollow, a high-risk niche or a site Google does not show can lower it. A site whose free listings are nofollow but which gives a followed link through a paid tier, for its badge or free is priced for that followed link once we have confirmed the route from its own pages, with the nofollow price beside it; a dofollow link only for a contest's winners does not count.

Level A — Counted

An editorial citation: the site linked because it is useful to its readers.

Level B — May be discounted

A guest, sponsored or placed article, or a product or tool directory listing.

Level C — Usually neutralised

A press release or syndicated copy, a forum post, profile, footer or sidebar link — and any nofollow link.

Level D — Network

A site on a server whose sites show footprints of one owner, a link farm, or machine-generated pages.

The same site is worth very different amounts depending on the link it gives you, so each report shows the value at every level.

How to read a report

What every site report's figures mean, and what they cannot tell you, said once here instead of on each report.

• BV is an estimate of ranking value, not a verdict on the site or a price anyone charges, and it does not predict what a link will do for a particular site.
• Relevance is neutral (×1) on a report. In an audit it runs from ×0.5 for an unrelated site to ×1.25 for the same topic.
• Site Strength runs from 0 to 100 in steps of 5: half the site's authority rating, half its organic traffic on a log scale; 100 is the strongest site on the web.
• Worth up to is a followed editorial link at the site's strength, before relevance.
• Traffic figures come from a third-party traffic source, named with its date on each report; no one outside a site sees its analytics.
• Domain age is a fact from the registry record, not a measure of a site's quality. BV does not price on age, but a domain registered in the last 30 days is never priced as not in Google.
• A followed link is not necessarily a link search engines value; a report does not measure rankings.
• Counts of sites, networks and hubs are mxverify's own, taken from its directory at the last daily update. Each pill on a report opens every site with a report that shares it.
• The badge a site can embed is a plain link and image: no script runs on the site that shows it.

How a site is classified

Every site with a report is in one of three states, and a site whose owner opts in also carries a fourth mark. The state is shown above the report's heading.

A site is audited when links placed on it have been checked by a bklink audit: each page fetched, the link found or not, followed or not, where it sits. A site is valued when someone has looked it up on the backlink value checker. Valuing a site includes a reading of the site itself — its homepage and up to ten articles: the newest few from its feed, the rest spread across its sitemap, so a reading reaches back beyond one week's posts — recording how those articles link to other sites (followed, nofollow, sponsored or user-generated), whether they say they are sponsored or contributed, and whether the homepage offers guest posts, sponsored posts or advertising. A model then judges the kind of links the site gives, from those observations only. Each report grades its reading: strong with six or more articles and thirty or more links to other sites, spread over at least two months where the articles are dated; thin below three articles or ten links; fair between. The reading is refreshed at most every 30 days. A valued site's record is offered to search engines once it carries that reading; it is counted as a publication only once it is audited.

Backlinks audited

A site is backlinks audited when its own owner audited the links pointing to it with bklink, and chose to publish that here. Its report says so, with the month of the latest audit, how many of its backlinks were checked and what share were live — every link in that audit, placed or scan-found, since together they are the site's backlink profile. It is a different thing from audited: audited is about links placed on a site; backlinks audited is about links pointing to it. What it changes about the site's link from mxverify is set out in Which links on mxverify are followed.

Topics

A publisher's topics are what it writes about, read from its own sitemap. Up to 40 of the articles it names are taken, spread across the list so they reach back in time, and each one's title is read from its address and filed under one topic in a fixed list of 16 subjects and 69 subtopics (Technology › AI, Sports › Cricket, Gaming › Puzzle & word games, and so on). Pages in a sponsored or press-release section are counted apart and never filed under a topic; the report says what share of the site they are. A site belongs to a topic when at least 15% of its sampled articles, and at least 4 of them, are on it; a subtopic's articles count toward its subject too. Only sites mxverify reads as publishers have topics.

Each topic pill on a report opens the list of publishers covering that topic, ranked by topical strength: the topic's share of each site's sampled articles, counted by mxverify. It is not a ranking of the web, a traffic figure or an authority score, and BV does not price on it. A subtopic with fewer than 10 publishers opens its subject's list instead. Titles are read, not whole articles, so a site whose addresses carry no title words cannot be read, and a site is read again about every three months.

Named editorial staff

A publisher's report says whether its own about, team, author and write-for-us pages name editorial staff: editors or authors with roles or profiles, or a masthead. Those pages are found from links on its homepage, read, and judged by a model against four plain descriptions, from "no named people" to "a masthead naming several editorial staff". The report shows the result as found, partly (a named owner or generic author names only) or not found, with the pages it read, never as a number: on the cases its readers disagreed about, the model put a site in the right group 13 times in 23. A site whose pages could not be read says so. BV does not price on it.

How a person checks the models

Some readings on mxverify are a model's: what a site is for, its topics, whether it names editorial staff. Before such a reading is published, and whenever it changes, the maintainer checks it in a review queue: a random sample of sites, and every site where two independent readers disagree. Each answer does three things. It measures the model, so what is published says how often it is right. It decides what is published and how: a reading that fails a check is withdrawn, reworded or narrowed. And it corrects the site it is about, permanently: a correction is stored apart from the model's reading and no later reading undoes it.

Where a person decides, and what each decision changes:

What the reviews of 28 October 2026 changed:

• Site type. 103 sites labelled by hand. Every press-release subdomain (pr.site) the model had read as a publisher, 36 of 36 labelled, is a press-release feed, so a fixed rule now files them under Syndication instead of the model.
• Topics. All 52 topic lists reviewed: 47 kept as they were, and 6 sites taken out of topics they do not cover. 30 sites' topics checked: 28 right, 2 partly right.
• Named editorial staff. 28 sites where two readers disagreed scored by hand; the closer reader was kept, and the result is shown as found or not found rather than as a score.

What the reviews of 29 October 2026 changed:

• Site type, category and placed posts. 93 sites drawn at random across every site type and category, most worth $10 or more a link, labelled by hand. The site-type model was right on 61 of them, a second model on 67; the keyword rule that names a site's category, on 57. From the labels came five rules, each scored on them: an address beginning ipv4. or smb., a subdomain rated below Domain Rating 1, or one of several sites redirecting to the same site, is a link network; a site the site-type question leaves untyped takes the kind of site its reading found. With them the site type is right on 70 of the 93, and 481 sites were filed again.
• Dofollow routes. All 133 routes found on directories' own pages answered by hand: 90 kept as read, 43 corrected — most often a "free" route that was a paid one — and 1 rejected, a list of other sites' links. Only these answers price a link (BV 2.9); 77 sites' value rose through a confirmed route, firsto.co's from $21 to $102 a link.
• A site's level. mxverify.uk, this site, ruled level B. The pages our reading samples are its site reports, whose links to the sites they describe are nofollow unless that site has paid for a followed one, so the reading priced a new link at C, $7.56. What a site gets here is a followed, listed link on a page about itself: level B, $37 a link.

A person's answer is evidence about the sites checked, not a guarantee about every other site: the published accuracy figures are for the sample and the disagreements reviewed.

Placed links and scan-found links

A backlink scan finds every page linking to a site, including mentions nobody arranged: a directory's rotating "similar products" box, a list that changes every week, a page that quoted the site. Those links are checked like placed ones, but when one disappears it says nothing about the publisher — the box moved on, as it is built to. Counting them as placements made a directory read "0% kept" for links it was never given, while the listing actually placed there was still live. So since 3 October 2026 every figure about a publisher — its audited record and mark, the Top lists, hub and network pages, and the totals on the audits page — counts placed links only. Scan-found links are shown on their own, on the audits page. A site's BV score still reads the link types of every judged link on it, placed or found, since both are links the site carries.

Which links on mxverify are followed

Most links from mxverify to other sites carry rel="nofollow": a report names a site, it does not vouch for it. These are the only links to other sites that do not, and why.

• A backlinks-audited site's own "Visit" link. On the report of a site marked backlinks audited, the one link to that site — "Visit <site>" next to the report's date — is followed. All four of these must hold:
  1. The site's owner ran a bklink audit of the links pointing to that site. bklink audits are a paid service.
  2. The owner chose to publish it, on the audit's campaign page, agreeing to the site being named with its figures.
  3. The owner proved the site is theirs, with a <meta name="bklink-verification"> tag on its home page, which we fetch and read when they ask us to verify it.
  4. Their latest completed audit of the site is under 90 days old.
  The opt-in and the audit's age are checked each time the page is rendered: when the owner withdraws, or 90 days pass without a new audit, the link goes back to nofollow at the page's next render. It is one link, to that site only: the sites in its audit, the ones linking to it, gain nothing, and any other link to it on mxverify stays nofollow. Opting in changes no figure: not the site's BV score, not its audited record. It does keep the report open to search engines while it holds, whatever else the report holds, since a followed link on a page search engines may not read would mean nothing. No listing and no score on mxverify is sold. The one followed link that is sold is set out next, and its report says it was paid for.
• A paid "Visit" link. Since 24 October 2026, the "Visit <site>" link on a site's report can be made followed for a one-time payment ($9 at launch), with the report's "Dofollow link" button, paid by card through Stripe. The rules:
  1. No proof of owning the site is asked for. It is the same one link as above, to that site only.
  2. It lasts for as long as the report is on mxverify, unless we revoke it (after a refund, or for abuse); the link then goes back to nofollow.
  3. It is not offered to adult, casino or CBD sites, and a site later read as one of those loses it.
  4. The report says the link was paid for, and when.
  5. Paying changes no figure: not the site's BV score, its readings, its link level, its traffic or its audited record. Those come from the same data and rules as every other site's.
  Like a backlinks-audited link, it keeps the report open to search engines while it holds, since a followed link on a page search engines may not read would mean nothing.
• bklink.uk. mxverify is maintained by bklink's founder and its figures come from bklink's audits, so its links to bklink.uk are followed.
• The maintainer's own profiles on X and LinkedIn, marked rel="me", on the about and methodology pages, and on Mastodon, in every page's footer.
• "Listed on" badges. The home page shows the badges of directories that list mxverify. Each is the directory's own embed code, kept as the directory wrote it — some ask for a followed link back, and a directory may check for it. They appear on the home page only.

Domain age and category

Registration dates come from the public registration record (RDAP). Where a registry does not answer our lookup, or has no RDAP service, the date comes from a commercial registration lookup that reads the same record (or its WHOIS where there is no RDAP). Content categories are our own classification, read from each site's own page metadata; a report's "Also about" lists the further categories the homepage's title and description name.

Tech stack and registration record

Each site's homepage is read as it is served. The platform in front of it comes from the headers only that platform sends; the ad and affiliate networks, from code that belongs to one company alone — an ad tag's host, an AdSense publisher id, an affiliate script or tagged link — so a name is never a guess. The same reading notes the schema.org types in the page's structured data, whether it marks up its authors, and whether it links its own about, editorial-policy and contact pages; it is the homepage only. The ad load is a web technology index's count of ad networks live on the site, from the ad code it loads and the sellers its ads.txt file names: ten or more is ad-heavy, and a site with none is shown as such even when an ads.txt file or an advertiser's tag is there. The registration record adds when the registration expires, when it last changed, and any status a buyer should know: on hold, lapsed, or pending deletion. A registration can be renewed at any time, so an expiry is called near or passed only when the record was read close to the date; otherwise the report gives the day the record was read. A site the Internet Archive saved well before its current registration began was most likely dropped and registered again. None of these moves BV: they describe the site, they do not price a link from it. Each technology and each ad-load band has its own list of sites, which the report's pills open.

Links to a site

A report also shows the links other sites have made to it — not the links placed on it that bklink audits. How many sites link to it (registered domains, so a site's subdomains count once), how many links, how many of those sites link nofollow, and how many IP addresses and subnets the links come from come from a third-party backlink index, read again every month so the report charts the count month by month; many sites linking from few addresses is a sign of a link farm. The index's spam score of the links is the average spam score of the pages linking to the site. "Where the links sit" is where on the linking pages the index found them: in the article or main content, or in the header, menu, sidebar or footer. It describes placement only: a footer link is not proven to repeat on every page, and a link in an article is not proven to be earned. It is shown from 100 links, and not where nine in ten or more sit in no section the index could name. "Pages in Google" is the count Google gives for a site: search, its own estimate, which can be far off either way. Where Google gives no count but lists a full first page of ten, the report says "at least 10": a floor set by how many results we ask for, not an estimate, and such sites usually have far more. No authority score or rank is shown: BV's one authority input is Ahrefs Domain Rating, which is not shown here. None of these moves BV but "Pages in Google": when two site: searches at least 7 days apart both find no page, and nothing on record shows Google indexing the site — an audited page found indexed, organic traffic of 100 or more visits a month measured in the last 30 days, or a domain registered in the last 30 days — a link from it is priced at most level C. A search that failed, gave no count or listed a full first page never counts as finding nothing.

Top lists

Each Top list is one set of publications — all of them, or one category, domain ending or registration era — with three or more links audited by bklink, from every account's audits. It is one table with every measure as a column, and a visitor sorts it by any of them; nothing is ranked for them. Rates are shown from three links, with the sample beside them, and sort by their Wilson lower bound (95%), what the rate is at least given its sample. "Followed" and "Still in place" read each link's latest check; "Kept over time" compares a link's first check with one 14 or more days later; "In the article", "Editorial", "Outside links" and "Article length" come from Jev's reading of the placed page; "Original" is the share of placements whose article ran on no other site; "Registered" is the public registration date of registered domains (a subdomain carries its parent's date, so it is left blank). "First archived" is when the Internet Archive first saved the site's homepage, which a subdomain has of its own. "Link-graph position" is where the site sits among all hosts in Common Crawl's web graph by harmonic centrality (1 is the most central), read from the top 20 million hosts of the latest release; below that it shows "20M+". It measures how connected a site is by links, it is not an authority score, and BV does not use it. "Load time (LCP)" is how long the main content took to appear for three in four real Chrome visitors over 28 days (Largest Contentful Paint, 75th percentile), from the Chrome UX Report (Google), licensed CC BY 4.0; the site report adds INP and CLS, whether the site passes Core Web Vitals, and the share of Chrome page loads on phone, desktop and tablet in the same 28 days (a device with too few loads for Chrome to publish is named as such, not as having no visitors). "BV per link" is the site's BV, the same figure its report shows, priced from its record once a day. The other columns are our own measurements only: no authority rating, no vendor's traffic figure. A category, ending or era gets a list once it holds ten publications.

Limitations of these measures

• Audits describe the placements bklink customers chose to audit, not a random sample of any site's links. A link is judged as each check found it; a link removed and put back between two checks is not seen.
• BV is a model's estimate of what one new followed link is worth to rankings. It is not a price, and it does not predict what a link will do for a particular site.
• Ahrefs answers 0 for a domain it has no data on, the same as for one it has never crawled, so 0 is read as no rating: the site is valued at $0 and its report is kept out of search engines' index until Ahrefs rates it. A report is also kept out while it holds fewer than four of its eight readings (rating, traffic, audited links, our reading of its pages, registration date, tech stack, page speed, sites linking to it); a daily check puts it back once it holds enough.
• A shared server is evidence of common operation, not proof. Category, CMS and tech stack are read from a site's homepage and server headers alone.
• Link types are Jev's reading of each page, a model's judgement. Traffic estimates are a third party's, shown with their date.

What is not published

• Third-party scores. Ahrefs' Domain Rating is an input to the BV score and is not shown. Third-party traffic estimates appear on site reports, with their date. Their top search keywords also open keyword pages, which list A–Z the sites where a keyword is a top search keyword; no page ranks sites by those figures.
• The identity of anyone who ran an audit, and the sites their links point to — unless a site's owner publishes their own site as backlinks audited. Even then, no account, campaign or customer name is shown: only the site, the month of its latest audit and its figures.
• Article titles and anchor text.
• A model's raw scores: a newsroom reading is shown as found, partly or not found, not as its number.

Changes to this methodology

• 29 October 2026. How a person checks the models now lists every decision a person makes and what it changes, with the reviews of 29 October: site types filed again by five rules, and every dofollow route confirmed by hand before it prices a link (BV 2.9), and a person's ruling on a site's level, first used for mxverify.uk.
• 28 October 2026. Added topics and the topic lists, named editorial staff, and how a person checks the models. Press-release subdomains are filed under Syndication by rule.
• 24 October 2026. Added the paid "Visit" link: a site's report link can be made followed for a one-time payment; nothing else on the report changes.
• 3 October 2026. Figures about a publisher count placed links only; links a backlink scan found are shown apart (why). Added the backlinks audited mark and its followed "Visit" link, and listed every kind of followed link on mxverify.

Who maintains this

mxverify is maintained by Palash Bagchi, founder of bklink (Immortal Reality PA LLC), who wrote this methodology and the rules the pages are generated from. He is on X and LinkedIn.

Corrections

Publications change. A link that was removed may be restored; a site may move hosts. Figures update as new audits run. To report an error, write to support@bklink.uk with the page address.