WebScore

New — The dashboard now opens with a chart of your last 30 days: scans you ran and websites you added, day by day, with the totals beside it. Hover any day to see its numbers.

v2.1.92 — released October 1, 2026

Is it possible for you to change the way I login to my account?

Requested by Gabriel, 2 votes.

From a public review by Yusuke Matsuba (LandingBoost, https://x.com/yusukelp), posted here on his behalf — he did not file it himself.

In his words:

"Some of the recommendations can still feel a little technical, especially for non-technical founders. It would be even better if the issues were prioritized more clearly by expected impact, so users immediately know which fixes will make the biggest difference."

Two things in one: rank findings by expected impact so the biggest win is obvious at a glance, and soften the wording for founders who aren't developers.

Requested by the community, 2 votes.

Fixed — Menus and dialogs could open white on the black app. If you had ever set the theme to light, everything that pops open — the navigation drawer, the scan dialog, dropdowns, the site switcher — kept rendering light even though the product itself went dark. There is no theme switch any more, so there was no way out of it. They now follow the product rather than the old setting.

Improved — The mobile navigation drawer is the same design as everything else: hairline rules between the links instead of floating pills, square corners, and the accent mark on the link you are on.

Improved — The website menu inside the app uses one highlight for the page you are on, not two. The top section marked it with an accent rule and the section below it with a filled green box, which made one menu look like two stacked on top of each other.

Improved — The scan dialog was six colours — a different one per module — plus a gradient bar. It is one accent on a ruled list now: each check reads queued, running or done, and the only colour left is the score at the end, where colour actually means something.

v2.1.76 — released September 19, 2026

Fixed — Scores, metric readouts and code snippets are back in a monospace face. They were meant to be all along, but the font they pointed at had been removed, so every one of them quietly fell back to the body text — which is why columns of figures never lined up.

Improved — The preview on the homepage shows the report as it looks today. It was still a picture of the old design.

v2.1.75 — released September 19, 2026

Improved — WebScore looks like one product now. The marketing site, the app, the demo, the PDF report and the trust badge were five different designs; they are now one — a dark drawing sheet with square corners, hairline rules and a single emerald. The old look mixed rounded cards, drop shadows and a different accent colour per section, which is what made the site and the app feel like separate things.

Improved — The product is a website health check, not a website audit. An audit is a one-off thing done to a big site by somebody else; a health check is quick, repeatable and yours to run. The homepage title, description and headline say so now.

Improved — One emerald instead of eleven. Every green in the interface — and there were eleven different ones — is the same colour. The reds, ambers and greens that grade a score are untouched, because those are the one place colour means something.

New — The dashboard sidebar folds. Collapse it to a strip of icons when you want the room back, and it stays that way next time you open the app. Websites and their tracked pages moved into a full-height column of their own, so a site with a dozen tracked pages no longer scrolls inside a few hundred pixels.

Improved — The scan date selector moved to the right of the top bar, where there was empty space, instead of hanging off the end of the breadcrumb. Which scan you are reading is a property of the whole page, not a step in the trail.

Improved — A tracked page's report now reads in order: how fast the page is, what it is made of, how it is found, how machines read it, then what to fix. Core Web Vitals sit directly under the score they are made of, rather than below four cards of metadata.

Improved — PDF reports are one document again. The cover was dark and every page after it was white; the whole report now uses the product's own palette, with square panels and the fix panels legible rather than dark green on dark green.

Improved — The trust badge was redrawn: square, one hairline, flat fills, no gradients. It keeps both light and dark versions, because it is the one thing that renders on somebody else's page and has to sit on their background.

Improved — Notifications appear at the bottom centre of the screen, over a blur that fades away upwards, and they clear the bottom bars that phone browsers put over the viewport.

Fixed — Generating a PDF used to say nothing for the several seconds it took, then announce that the report had been downloaded — while the browser was still opening its save dialog. It now tells you it is generating, then that the report is ready and your browser will ask where to save it.

Fixed — Text no longer touches the vertical rules that frame every public page. Headings, paragraphs and tables sat exactly on the hairline at every width.

Fixed — Contrast: the default button in the product put white text on the bright accent at 1.9:1 — below the readable minimum, on every button of that kind. That, the severity badges on Analytics, the WCAG conformance badges, the billing status pills and the selected item in the website sidebar (which rendered its label in its own background colour, invisible) are all fixed and measured.

Fixed — Non-production copies of the site can no longer be indexed by Google. A staging clone was serving the same title and content as the real site and competing with it for its own searches.

v2.1.74 — released September 19, 2026

Improved — The homepage makes one argument now instead of nine. The module grid, the three-step flow, the leaderboard preview and the comparison table are gone, replaced by a single editorial run-through: what each part of a scan actually changes for you, followed by the workflow that turns a finding into a verified fix.

Improved — The hero is centred again and the interactive demo sits inside it rather than behind a button. Under the form it now shows the people who joined most recently and how many have joined in total, in place of the floating card of score improvements.

Improved — Typing a domain on the homepage no longer opens a full-screen scan panel with its own reachability checks. The field stays where it is, and creating the account is the next step — whatever you typed follows you through sign-in and starts the scan on the other side.

Improved — The live demo is a real report now rather than the app in an iframe: the same "fix this first" card and score history you get on your own site, its own loading states, and module links that behave like links.

Improved — The dashboard sidebar is flatter — one list of sections instead of a rail plus nested menus, with the account menu at the bottom where you expect it.

Improved — Every email WebScore sends was rebuilt on one layout: the real logo instead of an inline icon, a single light design that renders the same in every client rather than a dark variant that some clients half-applied, and consistent headings, lists and buttons. Sign-in and account-deletion emails now use it too.

Fixed — Lifetime access is now also granted when you land back from checkout, not only when Creem’s webhook arrives. A slow or missed webhook used to leave a paid account on the free plan until the retry landed.

Improved — Lifetime is simply $29.99. The founding-spots counter and the struck-through $49 are off the homepage and the pricing page, and the pricing page no longer carries a credit-pack section — packs stay where you buy them, in Settings.

Improved — Credits are just Credits now, not "AI credits": the same balance covers AI Visibility answers and Conversion audits, and Settings says so.

Improved — Domain Metrics no longer has a refresh button that could not refresh anything. The numbers update when you run a scan, and the card says when they were last updated.

Fixed — Retired credit packs are still recognised, so a webhook or refund that arrives late for a pack we no longer sell is still applied to the right balance instead of being ignored.

v2.1.73 — released September 18, 2026

New — Every scan now includes an AI Visibility report. It checks the four things that decide whether an answer engine can cite you: whether AI crawlers are allowed to fetch the page, how much of your content exists without JavaScript (crawlers don't run it), whether your content is shaped so an answer can be lifted from it, and whether the markup says who you are.

New — A no-JavaScript check that catches the most expensive AI-search mistake there is. If your page renders in the browser, GPTBot and ClaudeBot see a blank page — the report shows exactly what share of your words reach them.

New — Answerability scoring: which of your headings ask a question, whether a direct 20-to-120-word answer follows each one, and whether the page opens with the answer instead of a warm-up.

New — Ask the AI directly. Enter the questions your buyers ask, and WebScore puts them to Claude with live web search, then records whether you were mentioned, whether you were cited with a link, where you ranked in any list of recommendations, and which competitors showed up instead.

New — Prompt suggestions are free: WebScore reads your site and drafts ten buyer-style questions that never name your brand, which you can edit before running a check.

Improved — PDF reports now include an AI Visibility section: crawler access, how much of the page survives without JavaScript, how many question headings get a direct answer, and a score for each of the four areas, followed by every finding with its fix. It appears on the cover, in the executive summary and as its own section, and does not move the overall score — same as in the app.

Improved — The before/after proof card in the homepage hero now shows AI Visibility too: whether AI crawlers are allowed, how much of the page survives without JavaScript, and where the site stands on answerability, entity schema and trust. It reads as a current-state score rather than a gain, because the module is newer than these sites' first scans and there is no honest before to compare against yet.

New — A proper pricing page at /pricing. Every difference between scanning free, unlocking a $3 report and paying once for Lifetime, laid out line by line across scanning, the report itself, staying fixed and the AI tools, plus what a scan checks and the questions people actually ask before paying.

New — Clicking the URL field on the homepage now opens a full-screen scan panel: one oversized input, nothing else on the page competing with it, and the three things people want to know before typing a domain, that the account is free and previews every module score, what Lifetime costs, and the 14-day money-back guarantee. Submit from there and the reachability checks run in the panel rather than dropping you back to the page.

Improved — The scan panel can be left at any point: a close button in the corner, the Escape key, a click on the backdrop, or the link under the form. Whatever you typed is still in the field when you come back, and the page behind never scrolls while it is open.

Improved — On phones the before/after proof card shows just the categories now: Accessibility, SEO, Performance, AI Visibility and their scores. The per-check detail underneath each one was turning the card into a wall of small text at that width, so it waits for a screen wide enough to read it. Nothing changed on desktop.

Fixed — The Uneed and Product Hunt badges wrapped onto two rows on phones. They now stay on one row and scale down together, sized by height so they keep the same height as each other and their proportions, down to a 320px screen.

Improved — The homepage headline is now just "Website health audit", with the promise underneath it in one line: find what is holding back your traffic, conversions and rankings, and exactly how to fix it. The module list and the MCP mention that used to crowd the subheading moved out, since the MCP section is now the very next thing on the page.

Improved — Every em dash is gone from the landing page copy. Thirty-four of them, rewritten as real sentences rather than swapped for another punctuation mark.

New — The MCP server gained five tools and now runs as a loop rather than a set of one-off lookups. get_fix_plan ranks every open issue by how many score points fixing it recovers, so your AI picks what to work on instead of guessing. compare_scans diffs two scans and reports what was resolved, what is new and what is still open, so it can check its own work after deploying a change.

New — AI Visibility is now available over MCP. get_ai_visibility returns crawler access, how much content survives without JavaScript, answerability, entity schema and trust signals, and get_ai_visibility_brief hands your coding agent a brief written for the job: the failing checks, the change to make, what must not regress, and how to verify it. get_scan_history rounds it out with recent scans and their ids.

Improved — Scan results now lead with a short summary before the JSON: the score table, the issue counts and the five fixes worth doing first. An assistant that reads only the top of the response still gets the part that matters.

Improved — The fix rankings use the same deduction tables the analyzers score with, so a "+25 SEO" is the real delta. Performance and Best Practices are scored by Lighthouse rather than from an issue list, so their findings are marked unpriced and ranked by severity instead of being given an invented number.

New — Every blog article now ends with a real, visible FAQ section. All 37 have at least five questions, each answered in a single paragraph short enough to be lifted as a snippet, and the questions carry their own anchor links so an answer engine can cite the exact one.

Fixed — Eighteen articles were publishing FAQPage structured data whose answers appeared nowhere on the page. Google requires that text to be visible, and an answer engine reading the rendered page saw nothing at all. The FAQ frontmatter is now the one source for both the visible section and the schema, so the two cannot drift apart.

Improved — Nineteen articles had no FAQ at all, five had only three questions, and four opened without answering the question in their own title. All of that is fixed. Across the blog the number of question headings with a direct answer underneath went from 52 to 243.

Improved — The MCP server now sits directly under the hero, rebuilt as a single panel: what it does on the left, the paste-ready config on the right with a copy button and syntax colouring, and the tools it works in underneath. It replaces three stacked centred blocks that took twice the height to say the same thing. The hero subtitle now names both halves of the AI story too: AI visibility runs on every scan, and MCP is how an agent applies the fixes.

Improved — On the pricing page, Lifetime sits in the middle where the plan you should pick belongs, and each card now ends where its own content ends instead of being stretched to match the tallest one. The Creem payment badge moved below all three cards, since it covers every purchase and not just Lifetime.

Improved — The pricing block on the homepage is now just the decision: what you get free, what $3 unlocks, what Lifetime costs. The full feature lists moved to the pricing page rather than crowding the page you were still deciding on.

Improved — The homepage now leads with what each part of WebScore gets you rather than what it checks. Every section, from the three-step flow to the audit checklist, states the outcome first and keeps the technical detail underneath it.

New — The live demo now includes AI Visibility. It is a real report from a real scan, with crawler access, the no-JavaScript reading, answerability, entity schema and trust signals, plus the per-section "Copy fix for LLM" briefs.

Fixed — The AI Visibility issue list was titled "SEO Issues" and told you to improve your rankings, because it reuses the SEO component. It now says what it is, on both the dashboard and the demo.

Improved — The hero is lighter: the embedded demo frame is gone, replaced by a "Try the interactive demo" button that opens the full annotated report at /demo. Same thing to look at, without a second app booting behind the fold.

Improved — Landing page design pass: the hero is down to one headline, one input and two buttons, on a plain band with a single brand glow that now renders in light mode too. The announcement badge, the inline score line and the spinning gradient around the URL field are gone, and the "Try the interactive demo" button now has a real fill in light mode instead of reading as bare text. Alongside it, one heading scale that steps down on phones instead of six different sizes, shorter paragraphs, consistent icons on every card, and hover states that lift the card and nudge its arrow.

New — AI Visibility now knows what kind of page it is looking at. A landing page is no longer marked down for having no byline, no TL;DR and no citations — those are article conventions. Product pages are checked for Product schema with price and availability, listing pages for ItemList, and articles keep the full provenance checks. The detected page type is shown next to the score, with the reason on hover.

New — Every AI Visibility section has a "Copy fix for LLM" button. It writes a brief for your AI coding agent: the exact checks failing, why each one costs you citations, the change to make, what must not regress, and how to tell it worked. There's a whole-report version too, ordered by how much score each section is losing.

Improved — Checkout now takes the buyer from your signed-in session rather than the link, so a purchase can only ever credit your own account.

v2.1.72 — released September 10, 2026

Improved — The homepage now leads with what WebScore is: a free website audit. New title, headline and a "What a website audit covers" checklist that lists every check the scan runs, with the thresholds it holds your site to.

New — Three new guides: How to Audit a Website (the complete checklist with a copyable template), Website Audit vs SEO Audit, and Website Audit for Small Business, including what an audit should cost.

Improved — The live demo is now an annotated example audit report: each module explains what its score means and shows the top fix, and every module tab points Google at the one /demo URL.

Improved — The SEOptimer alternatives guide now compares seven tools with verified pricing and answers "is there anything better than SEOptimer?" directly. The old templated alternatives page redirects to it.

Improved — Refreshed the conversion audit tools, domain metrics and technical SEO audit guides with direct answers to the questions people actually search.

Fixed — Removed a hardcoded star rating from the site's structured data. Ratings in schema must come from real user reviews, and the site does not collect them yet.

Improved — The homepage document is 46% smaller (556 KB to 300 KB, 85 KB to 52 KB compressed): landing sections and the header's tool menu no longer ship twice, repeated icons come from one shared sprite, and the footer badge strip is sent once. Every page that shares the header and footer got lighter too.

v2.1.71 — released September 9, 2026

Improved — The Websites page now marks every website that is connected to Google Search Console with a "Search Console" pill next to its monitoring badge, so you can tell at a glance which of your sites have search data flowing into their SEO page and which still need connecting. Hover the pill to see the connected property.

v2.1.70 — released September 9, 2026

New — Monitoring now emails you when a scheduled check cannot load your website at all. Until now a site that went down failed its daily check silently — the report email only goes out for checks that complete, so the one day you most needed to hear from us was the day you got nothing. The notice says what the check saw, what it most likely means, and what to do, and repeats at most once a week while the site stays down. Your normal report resumes as soon as the site loads again.

Improved — When a scan cannot load a site over HTTPS, it now also tries plain HTTP and says what it found. "The server is up but its HTTPS setup is broken" points you at the certificate; "plain HTTP fails too" tells you the site is down. Before, both cases read the same.

v2.1.69 — released September 9, 2026

Fixed — Scanning a site whose SSL certificate does not match its domain, whose domain name does not resolve, or whose server refuses the connection failed with a puzzling browser message ("Unable to retrieve content because the page is navigating") — and now and then it went the other way and completed with scores computed from the browser's own error page. The scan now stops straight away and says what went wrong, for example "Could not load https://example.com: the site's SSL certificate is not valid for this domain (ERR_CERT_COMMON_NAME_INVALID)", so you know whether to fix the site or just try again.

Improved — When the page cannot be loaded at all, the performance audit is no longer attempted against the browser's error page.

v2.1.68 — released September 8, 2026

Fixed — A domain deliberately set up to send and receive no email — a null MX record, an SPF record of "v=spf1 -all", a reject DMARC policy and a wildcard DKIM record with an empty key — was graded as if it were misconfigured. The wildcard DKIM record, which revokes every signing key, looked to the scanner like a DNS setup that answers every name, so DKIM came back "can't be verified". And the DMARC record was marked down for not asking for reports that a domain sending no mail has no use for. Both were false alarms and no longer appear.

Improved — The Email Authentication card now recognises a null MX record and says so: the domain is graded as a no-mail domain, DKIM shows "all selectors revoked" instead of "missing", and the fix instructions for any gap give the exact records such a domain should publish rather than a mail provider's SPF include.

Improved — The free Email Security Checker and the check_email_auth MCP tool make the same distinction, so all three give a no-mail domain the same verdict.

Improved — A DKIM record with an empty key is a revocation, not a key. It is no longer counted as DKIM being configured on a domain that does send mail, and a domain that revoked every selector while still publishing mail servers is told so, with the two ways to resolve it.

v2.1.67 — released September 8, 2026

Fixed — A Conversion audit could fail with "The AI analysis returned an unusable result" on pages that use emoji in headings, buttons or link text. When the audit shortened that text to fit its request, it could cut an emoji in half, and the half-character made the whole request unreadable to the AI. The text is now cleaned before it is sent, so an emoji at the cut point no longer sinks the audit.

v2.1.66 — released September 4, 2026

Fixed — Running a Conversion audit on a second website could fail straight away with a browser error ("Target page, context or browser has been closed") even though the first one had worked. The audit kept one browser open between runs to save a second, and when that browser crashed, every audit after it was handed the dead browser. Each audit now gets a fresh browser of its own, and if the browser crashes mid-capture the audit retries once before giving up.

Improved — The audit's browser now runs in the same, sturdier mode the site scanner already uses, so one heavy page can no longer take the whole browser down with it.

Improved — When the browser does crash, the audit now says so in plain words instead of showing the raw error.

v2.1.65 — released September 3, 2026

Fixed — Some sites came back with a Best Practices score of 0 while every other module scored normally. Nothing was wrong with those sites — the audit behind that one module had failed to run, and a failed run was being written down as a score of zero. Because a site's overall score is the average of its five modules, that phantom zero quietly cost it around twenty points of overall health as well.

Improved — Best Practices is now measured through Google PageSpeed Insights, the same place the Performance score already comes from. It runs on Google's machines instead of ours, so it no longer competes with the scan's own browser for memory, and sites that turn away traffic from our server no longer turn away the audit.

Improved — When the audit genuinely can't run, the module now says so and records no score at all, instead of showing a 0 that reads like a verdict on your site. Scan again and it fills in.

v2.1.64 — released August 22, 2026

Fixed — Scanning from the homepage could answer "that host is not allowed" for perfectly ordinary sites — google.com among them. The safety check that keeps the scanner away from private, internal addresses was reading every IPv6 address as internal, so any site reachable over IPv6 was turned away before we even tried it. Sites without IPv6 were unaffected, which is why it looked arbitrary: github.com worked, google.com didn't.

Improved — The same check now understands addresses that carry an IPv4 one inside them, so it judges the address a name really points at rather than the way it happens to be written.

v2.1.63 — released August 17, 2026

New — The whole release history — all 71 versions back to 2.0.0 — now sits on the public changelog hub next to the feedback board and the status page. You can link someone to a single release, or subscribe by RSS and stop checking the site to find out what changed. This page stays the place releases are written; the hub mirrors it.

v2.1.62 — released August 17, 2026

New — Feedback now goes to a public board instead of a private email to us. You can see what other people have already asked for, add your voice to it rather than filing the same thing again, and follow what happens next. The button sits bottom-right on every page inside the app, and there's a Feedback link in the site footer. If you're signed in, your name and email come along, so all you write is the idea itself.

New — A "What's new" badge on the public site opens the release notes in place, and shows a dot when there's something you haven't read yet. It stays out of the app itself — inside your dashboard the only thing floating is the feedback button.

New — The site footer now carries live uptime for the last 30 days, and links through to a full status page. If a scan ever fails on you, you can check in one glance whether it's us.

v2.1.61 — released August 17, 2026

Fixed — Scans run from our servers could report "TLS 1.3 not enabled" for sites that have it — the check parsed openssl's console output, whose format differs between systems and often omitted the line we read for TLS 1.3 connections. The scan now asks the TLS connection itself, which gives the same correct answer everywhere. If your security report claimed TLS 1.3 was missing, re-scan — it wasn't.

v2.1.60 — released August 16, 2026

Fixed — The free SSL checker flagged sites on TLS 1.3 with "weak cipher strength: 0 bits" and docked their grade — the newest, strongest protocol scored worse than the old one. The cipher strength is now read correctly for TLS 1.3 suites, so a modern setup gets the clean A it deserves. The full scan and dashboard were unaffected.

v2.1.59 — released August 16, 2026

Improved — The scan now discovers custom DKIM selectors instead of only checking a fixed list: names derived from your own domain (a real user's key lived at yourdomain._domainkey), rotation years like s2025, and your email providers' conventional selectors on their sending subdomains — with a wildcard-DNS guard so guessing can never produce a false positive. Found keys also report their strength, so a 1024 to 2048-bit upgrade shows up as exactly that. Thanks to the bepublish.com report that drove this.

Improved — Your AI assistant can verify any selector directly: the MCP email check accepts a dkimSelector parameter and reports precisely whether that name answers.

Fixed — Websites added in the last few days could fail their first scan with "Page not found for this website" — adding a site stopped registering its home page behind the scenes. New sites are created correctly again, and any affected site repairs itself automatically on its next scan.

v2.1.58 — released August 15, 2026

Fixed — Best Practices said "no service worker" on sites that have a perfectly good one — the check only looked once, at one moment, and many sites register their worker a few seconds after load (or skip it for scanner visits entirely). The scan now also waits briefly, looks for registration code on the page, and checks the usual sw.js locations directly — and the result says how it found yours. Thanks to the user who reported it.

Fixed — Two more Best Practices checks got more accurate: off-screen image counting no longer depends on how far the scan had scrolled the page, and the preconnect suggestion no longer counts links the page never actually loads (like alternate-language links or RSS feeds) as third-party origins.

v2.1.57 — released August 15, 2026

New — The Results page is now a collection of playbooks — real websites' improvement stories written from their actual scan history. Each one shows a scoreboard of measured before/after values, a chart of the score across the site's lifetime with markers naming what moved it, and numbered plays: what the scan found, what was done, what the fix buys you, and what skipping it costs. Every number is a measured value; nothing is estimated.

New — Each playbook links to the featured site's public Trust page — the playbook is the history, the Trust page is where the site stands today.

Improved — The navigation grew up: Results now has its own place, Tools became Free Tools, and its menu opens as a full-width panel showing every free tool with what it does — no more guessing from a shortlist.

New — A new comparison on the blog: WebScore vs OptimizeMyWebsite.ai — what each audit actually checks, honestly, including where the other tool is the better pick.

v2.1.56 — released August 15, 2026

Fixed — Our own accessibility scan flagged the homepage for low-contrast text — the scan consent line, the crossed-out old price, the Lifetime button and the before/after checklists. All of them now meet WCAG AA contrast, verified with the same check that caught them.

v2.1.55 — released August 14, 2026

Improved — A tracked page's report now previews the same way the website modules do. Free accounts see the page's scores, Core Web Vitals and a few of the smallest issues; the detailed analysis — page content, indexability, canonical, structured data, links, agent readiness and performance opportunities — unlocks with the full report, from the same $3-per-report or Lifetime access.

Improved — Best Practices keeps its findings for the full report. The free view shows a few of the checks your site passes; what failed, how serious it is and how to fix it are part of the paid report, with the unlock prompt counting exactly what's waiting.

Fixed — Typing your own domain into the homepage scanner no longer answers "that host is not allowed" — a server-side safety check was mistaking the site's own address for a private one.

v2.1.54 — released August 14, 2026

Improved — The homepage now answers near-instantly for visitors who aren't signed in — it's served from a cache instead of being rebuilt for every visit, and refreshes itself every few minutes. Signed-in visitors still get their live, personalised page at the same address.

Improved — The sign-up and sign-in pages shed about a third of their code. A whole validation library was being downloaded to check that an email looks like an email — the same checks now run without it, so the form is ready sooner, especially on phones.

Improved — Marketing pages stopped downloading the homepage's code. A quirk in how components were bundled meant even the privacy policy carried the full landing page with it; each page now ships only what it actually shows.

New — Everyone appears in the "Loved by" row on the homepage now: accounts that signed up by email and have no profile photo get a friendly generated portrait instead of being left out.

v2.1.53 — released August 14, 2026

New — Run an AI Conversion audit on any page of your site, not just the home page. Type a path like /pricing (or paste the full address) in the audit wizard and the AI reviews that page. Subdomains count as their own website — add them separately — and a mistyped address is caught before it costs you an audit.

New — Recommendations are now a working checklist. Mark items done or dismiss them, watch the fixed count climb on the report and in your history, and send any recommendation to your Tasks list to plan it alongside your scan findings.

New — Your score comes with a breakdown across seven dimensions — visual design, layout, typography, navigation, calls to action, accessibility and consistency. Re-audit the same page and the report shows what moved since last time, overall and dimension by dimension. Audits of different pages are never compared with each other.

New — Audit mobile and desktop in one go: the wizard's new "Both" option queues the two audits back to back. It uses two audits from your monthly allowance.

New — The "Do this first" card — the one change the AI judges most valuable — now opens every report, not just shared ones. The screenshot and the action plan are linked both ways: click a recommendation to see where it sits on the page, and click a marker to jump to its full write-up.

Improved — Audits run on the same queue as scans now, so a crashed audit can no longer block the next one. Failed audits stay in your history with a plain explanation and a Try again button — and they don't count against your monthly allowance.

Improved — You can delete an audit from your history, and its report and screenshots are removed with it. Share links now use cryptographically random addresses.

Improved — The AI reviews the full page alongside close-ups of each section and returns its findings in a strict format, so analyses fail less often and recommendations anchor to the right spot on the screenshot more reliably.

Fixed — The raw HTML copy of an audited page is no longer stored or reachable on the web — only the screenshots the report actually shows.

v2.1.52 — released August 14, 2026

Improved — Enter your address on the home page and we check the site right there — that it's up, how fast it answered, whether it's on HTTPS — before asking you for an account. Signing up then starts your scan automatically, and the sign-in and sign-up pages show a banner confirming a scan is waiting for you.

Improved — The leaderboard's header and closing call-to-action use that same scan form. If you're already signed in, entering a URL simply starts the scan.

Improved — Smaller touches: the pricing section marks the Lifetime plan as best value, and the home page shows how many users WebScore is loved by, right next to their avatars.

v2.1.51 — released August 14, 2026

New — Every article now has a table of contents that follows you as you scroll — beside the text on a large screen, collapsible at the top on a phone. Headings are linkable, with an anchor that appears on hover.

Improved — Code in articles got proper treatment: blocks show their language and a copy button, inline code reads as tinted chips, and keyboard keys, footnotes, checklists and collapsible sections all render the way they should.

Improved — A cleanup pass across every article — emoji removed and formatting made consistent.

v2.1.50 — released August 14, 2026