What shipped, dated

No releases in that area yet.

1. 4 Oct 2026

   HighRegard for WordPress 1.0.0: from a read-only dashboard to a reviewed action surface, with signed updates

   The plugin that began on 2 October as a read-only view of a workspace inside wp-admin reaches 1.0 after seven further milestones, each gated by the same test set and each reviewed before the next. It still opens on the dashboard people know (greeting, headline, four tiles, the list of things waiting, the 28-day view, what changed since the last visit), fed by the workspace document, the Website Score and read-only checks on the site’s own database, painting first from held data and loading the rest in the background. What is new is everything around that. Connection is a pairing: Connect to HighRegard sends the administrator to the app with a random state, the app approves and hands back a single-use code, and the plugin exchanges it server to server for a key bound to that one website; a key presented by any other host is refused. SEO fields are read through adapters for Rank Math, Yoast SEO and SEOPress (All in One SEO is detected and declared unreadable rather than guessed), and page text and links through Elementor, Divi, WPBakery, Bricks and Beaver Builder data as well as the editor, so a finding says which field it read. Every finding carries a fingerprint and an evidence drawer naming the fields, the sample and the time, and “Check the live pages” asks the platform to fetch up to ten of its pages as a crawler would and marks it rendered confirmed, contradicted or unknown. Findings can be decided (fix later, already fixed, accepted risk, not relevant), site-wide and reversible; a Change Guard records changes to the settings that decide whether the site can be found; the workspace’s anomaly detections appear as “worth a look”. A panel in the editor shows a page’s own 28-day search figures, the searches that bring people to it, whether it is losing clicks, pages that mention it without linking to it, and the telephone, WhatsApp, email, booking links and forms it carries. Prepare fix drafts a meta description or SEO title from the page’s own text, the business and its search evidence, shows it beside what the field holds now with the reasoning and the facts, validates it against those facts (length, no invented figure, no claim the page does not make) and labels it AI or rule-based. Only then, and only when an administrator has switched applying on, can a person with the manage capability apply that text to the page through the SEO plugin’s own field, after a two-click approval of the exact text; every change is recorded with its previous value and undone exactly, refusing if the field changed since, and once a full 28 days of search data lie after it the record shows click-through, clicks, impressions and position before and after, worded as observation with confounders listed. The platform has no write endpoint: a stolen read key cannot change a site, and the threat model for that was written before the write code. 1.0 adds the release machinery: a manifest on highregard.co.uk signed with Ed25519, verified by the plugin before it is parsed, checked against the server’s PHP and WordPress versions, then handed to WordPress’s own update flow, which gets the package only after its SHA-256 and size match the manifest; stable and beta channels, a previous-version reinstall through the same checks, and nothing offered at all without the sodium extension or an embedded key. The plugin states the document schema it understands on every request and the platform answers plugin_outdated or platform_behind rather than a document that would be misread; fixtures for every document come from the platform’s own serializers and a test on each side fails when they drift. Usage counts (whole numbers only) sit on the Settings screen and travel with the hourly heartbeat unless switched off; the workspace’s settings page shows each installation’s version, an update badge when it is behind, and which sites have gone quiet. The zip ships its documentation: threat model, privacy and data flow, API contract, compatibility matrix and support policy, updates and signing, security review, performance notes, observability and the release checklist. Forty-seven standalone PHP tests and one hundred and nine platform unit tests; the compatibility workflow for PHP 7.4 to 8.4 and WordPress 6.0 to current is written and waits for its first run before “Tested up to” moves.

   ApplicationWordPress
2. 22 Sep 2026

   Research note 2: 118 Yorkshire dental practice websites that rank, measured with the site’s own engine, dataset published

   The first study with a sample rather than a calibration. The top-20 organic results for “dentist in [town]” in Leeds, Bradford, Harrogate, Huddersfield, Wakefield, York, Halifax and Sheffield were collected on the evening of 22 September; directories and NHS pages were removed, and the 127 provider pages that remained were each submitted to the same /api/analyse endpoint the free checker uses, in site mode, engine v1.1.0. 118 could be fetched (one national chain’s pages returned nothing to the crawler; one site redirected in a loop). The note reports the distribution (median Website Score 86; 61% strong), where the marks are lost (structured data and search presentation), the structured-data gap in detail (22% no JSON-LD, a further 14% with JSON-LD but no business entity, 35% declaring the Dentist type), search presentation and conversion flags (19% no meta description, 14% no tap-to-call link), chain against independent pages, a by-town table with its caveat, and the finding that organic position has no measurable relationship with Website Score (Spearman 0.008). Every figure is computed from the published dataset at build time; the dataset is a CSV of 118 anonymised rows under a CC BY licence, with Dataset structured data on the page. Limitations are listed, including that a sample of what ranks says nothing about what does not.

   Research
3. 22 Sep 2026

   The founder entity: regulator register linked, credentials machine-readable, profile slots ready

   The author page now links the two independent records that verify who writes here: the General Pharmaceutical Council’s public register entry for the founder (registration number and independent prescriber annotation, on the regulator’s site) and the clinic that names him as founder and clinical lead. The same facts are declared in the page’s structured data: the Person node carries the registration number as an identifier, two credentials (the MPharm degree and the professional registration, each with the body that recognises it), the subjects he writes about, and sameAs links to the register and the clinic; the profile page names him as its main entity and the About page names HighRegard as its own. LinkedIn and Instagram profiles do not exist yet, so they are not declared: the code has slots for them that emit nothing until a real URL is added, and a test fails the build if a placeholder ever reaches the graph.

   SiteMethodology
4. 22 Sep 2026

   Eight service pages published, each bought through a product that is already on the pricing page

   Local SEO, Google Maps SEO, AI search visibility, schema markup, technical SEO, Google Business Profile, citation management and local landing pages each have a page under /services/, with an index. They were written in August and held back at launch because they quoted prices for products that were never sold; they have been rewritten so that no page carries a price of its own. Instead each has a “how you buy it” block naming the live products that deliver the work (the £295 audit, the £795 Repair Sprint, the £495 and £795 monthly plans, the Transformation) with the price read from the same configuration as the pricing page, and a test fails the build if a service page ever shows a figure that is not one of those. Each page also states what the work includes, what it deliberately does not, how it runs, where it fits across Google, Maps and AI answers, which free tools measure it, three related Insights articles, and the Google guidance and HighRegard measurements its claims rest on. The AI search visibility copy now matches the Growth plan’s written scope: at least twenty prompts a month against at least two named assistants. “All services” is in the Services menu and the footer.

   SiteServices
5. 22 Sep 2026

   Every free tool page now carries a six-line scope and a worked example produced by the tool itself

   Each of the twenty tool pages gained an at-a-glance block that says, in the same six lines, what you enter, what you get, what the tool reads, what it cannot see, how long it takes and, for the scored checkers, where the method is published. Under it sits a worked example. For eighteen tools the example is the tool’s own output on a fixed input, generated when the site is built: the five page-analysis tools run the real engine on the homepage of a fictional Harrogate dental practice written to have a typical mix of strengths and gaps (it scores 64, with the tap-to-call link, the meta description and the location in the title as the first fixes), the six calculators call the same functions as the page, and the generators use the same URL rules. Each example ends with a short reading of what the result means and what you would do next. The Live AI Visibility Test and the Domain Authority Checker, which need a live model or a live registry, show a sketch of a result labelled as illustrative, with nothing in it measured. A test now fails the build if any tool lacks a scope or an example, or if the fixture stops producing a mixed result.

   Tools
6. 22 Sep 2026

   Live now and coming next stated in one place; a shorter home page; the phone main-thread budget met

   What the application does today and what it does not yet had to be pieced together from a preview window, a footnote and the platform page. One block now states it, and it is rendered from a single source in three places: under the hero window, in the home page’s platform section, and at the top of the platform page. Six things are live in the demo and every workspace (Google Ads, Search Console, Analytics 4, Google Maps geo-grids, tracking health and the opportunity queue); three are coming and are not sold as software (AI visibility, reputation, authority and citations), with the free tools and the audit named as how they are measured in the meantime. The home page was cut from fourteen sections to eleven: the illustrated feature grid, the six-dimension list, the AI-search explainer and the four-step process now live on the pages that already covered them (audit, AI visibility, process), leaving the home page as a summary that links out. On the lab phone profile the home page’s total blocking time fell from about 720 ms to about 265 ms against a 400 ms budget, and its largest contentful paint from 2.1 s to 1.3 s: sections below the fold are no longer laid out until they are near the viewport, and four boot-time scripts that read element positions (and so forced the whole page to lay out inside the first task) now wait for the first frame or for an intersection observer instead. Every measured page improved; none regressed. One queued article topic and five others that would have duplicated the rewritten articles are marked skipped so the weekly pipeline does not repeat them.

   SiteApplicationPerformance
7. 17 Sep 2026

   Every Insights article rewritten as a long-form, referenced piece, at the same address and with its original date

   All eleven Insights articles were researched again and rewritten into the long-form format the weekly pipeline uses: a headline that states a claim, a standfirst, an at-a-glance box, sections that each answer one question, first person only where the founder’s own experience is the evidence, and numbered references with the date each was checked. They run to roughly 2,850 to 3,200 words each, against 500 to 800 before, and carry between 16 and 25 references apiece, 179 distinct sources in all, every one fetched twice before publication, once by the writer and once by an independent check. Each reference now carries an evidence label (platform documentation, regulator or government, study or research, statistics, reporting, or HighRegard measurement) so a reader can weigh it without following the link. URLs and original publication dates are unchanged; each page shows its update date. Where a source could not be verified, the claim was removed rather than kept: the piece on regulated clinics, for example, no longer quotes the current penalty wording of section 10 of the Health and Social Care Act because only the as-enacted text could be fetched, and says so.

   Insights
8. 16 Sep 2026

   A weekly long-form Insights article, referenced and checked before it is filed; the author page in the navigation; credentials on every byline

   Insights is now set up to gain one long-form article a week (the first is scheduled for Tuesday 22 September), written in the explanatory register of a broadcaster’s in-depth desk: a standfirst, an at-a-glance box, sections that each answer one question, first person where the founder’s own experience is the evidence, and numbered references with the date each was checked. The pipeline that produces it researches with live web search, fetches every candidate source and drops any that does not answer, writes to a fixed structure, validates it (length, every citation matched to a reference, every reference cited, no figure without a citation, no promises about rankings), has a second model check every cited claim against the text of its source, revises once on failure and files nothing on a second failure. The article is committed to the repository so the site rebuilds; a note at the foot of each one says how it was produced. Every byline now carries the author’s credentials (MPharm; Founder and Director of Pulse & Clarity Clinic; Founder of HighRegard) with an author box linking to the author page, which is now in the Insights menu and the footer alongside the three free guides; a test fails the build if any indexable page becomes unreachable from the home page.

   InsightsSite
9. 16 Sep 2026

   Domain Authority Checker: registration data now comes from each registry directly

   The free Domain Authority Checker was returning “unavailable” for every .uk domain because the public RDAP redirector it relied on refused the lookups. It now reads the IANA bootstrap file to find the registry responsible for each domain ending (Nominet for .uk, Verisign for .com, and so on) and asks that registry directly, keeping the redirector only as a fallback. The table shows the registration date, age and registrar, says “Not registered” when a registry confirms a domain is free, and no longer shows empty columns for authority providers that are not connected on this deployment; one note explains which scores are live and which are not.

   Tools
10. 14 Sep 2026

    The Google Maps module is live: geo-grid scans in the workspace, paired change, and Maps findings in the queue

    The application gained its first local-layer module. A HighRegard GridTracker scan (CSV or JSON export) is imported into a workspace as an immutable measurement with its grid fingerprint; the Google Maps page then shows the rank at every point as a grid with a compass, coverage by ring outward from the door and by direction with the weakest side named, per-term figures with the top-3 radius, a change map and paired deltas against the previous scan on the same grid (suppressed below 80% paired coverage, as the tracker’s methodology requires), the trend across comparable scans, and the competitors that hold the top 3 where the business does not when the JSON export is used. The overview carries the grid and the top-3 coverage; the opportunity feed writes gaps, absences and declines as measured findings with the ring and direction evidence attached; the data sources page lists the geo-grid as available. The definitions are the tracker’s own, so a figure in the workspace equals the figure on its report. The demo workspace ships with two synthetic scans, and the platform page, the home hero window and the new module page at /platform/google-maps/ are generated from that dataset. The composite score methodology now records Google Maps as a live module with its formula.

    ApplicationSiteMethodology
11. 14 Sep 2026

    One demo dataset everywhere, a Trust Centre, HighRegard Research, evidence labels, estimate classes, and a performance gate

    The demo workspace is now frozen at one date, and the platform page and home page are generated from that same dataset, so the figures quoted on the site are exactly the ones in the demo; a test fails the build if they drift. The home hero window is labelled as a platform preview with the forthcoming modules marked, and carries a plain-text equivalent for assistive technology and crawlers; decorative grids and duplicated previews no longer appear as text in the HTML. New: a Trust Centre written from the code (read-only scopes, per-workspace encryption, deletion on disconnect, audit log, what the AI provider sees); a Research section with its first note, the geo-grid calibration, and the planned studies listed without numbers; a draft methodology for the composite HighRegard Score; sources with evidence labels (Google guidance, measured, judgement, hypothesis) on the AI visibility page and every Insights article, and a one-click “preferred source” link for Google. In the application, every estimate now carries a class — measured, empirical, heuristic or scenario — with the reasons behind its confidence, and a language check removes repeated sentences from recommendations. The sitemap’s dates now reflect real content changes. A lab performance gate runs before each release; the hero fog is skipped on devices that would pay for it, and the site boots in two phases so first paint is not held behind the rest.

    SiteApplicationLegalInsightsMethodology
12. 14 Sep 2026

    A pricing rail: one card forward at a time, a travelling light on that card, and the last plain pages join the kit

    Pricing on the home page and the pricing page is now a dark rail of product cards. One card sits forward in the centre with a live lightning border and the white button; the others stand back, dimmed by distance. Select any card, use the arrows, the labels underneath, the arrow keys or a swipe to bring it forward. The active card carries a light pulse: two soft comets of light travel its edge a half-turn apart, each a long tail brightening into a white head, with a blurred glow following outside the edge and the card’s halo breathing slowly. It is pure CSS, and under reduced motion it rests as a steady tinted edge. Cards elsewhere on the site are plain again. The remaining pages moved onto the kit: the author page has a proper portrait section, figures, a working rail and an article grid; the changelog opens with the latest release, counts what has shipped and filters by area; the legal documents and the scoring methodology gained a sticky contents list that tracks your position, numbered sections and weighted bars for the seven score components; every free tool page opens beside an illustration, sets its explainer out as numbered cards and shows related tools with icons; the not-found page points the way with icon cards.

    SiteDesignTools
13. 14 Sep 2026

    The application moves: counting figures, sortable tables, a first-visit tour, and a real current around the priced cards

    Every app page now rises in as it loads and its figures count up; tiles and panels carry the site’s pointer spotlight; tables sort by any column and filter as you type, with the header pinned while you scroll; sparklines draw themselves; the tracking-health score fills as a ring; flash messages slide in and can be dismissed; the blocked-until-tracking-is-fixed tile breathes. The demo workspace offers a 30-second tour on first visit, and keyboard shortcuts (/ to filter, [ and ] to change the window, ? for the legend). On the marketing site, the priced cards now carry a real electric current: a bright dashed stroke running around the edge with a trailing glow, a counter-current of small sparks, and irregular flashes, each card in its own colour.

    ApplicationSiteDesign
14. 14 Sep 2026

    The application now looks like the site’s product surface

    The workspace and the live demo have moved onto the same black surfaces, Inter Tight figures, mono labels and pill controls as the product tiles on the marketing site, with the HighRegard mark in the sidebar and a window-style crumb above each page title. Tables scroll inside their panels on phones instead of widening the page, and the phone navigation is a row of pills.

    ApplicationDesign
15. 14 Sep 2026

    The remaining pages join the kit: case studies, sectors, insights, contact, snapshot, sample report, resources

    The case study opens with its headline figures counting up and the work as a numbered rail, with the limitations quoted before they are read. Each sector page carries icon cards, a statement band, the search-behaviour rail and the three products as charged cards. The insights index is a grid of article cards; contact and the free snapshot gained illustrations and icon cards; the sample report opens with its numbers and lists the remaining sections as a rail; resource pages gained a scan illustration. Also fixed: a long label overflowing narrow screens, and the integrity fingerprints on the case study forcing the page wider than a phone.

    SiteDesign
16. 14 Sep 2026

    Every priced card carries its own charge, and four fixes

    The free snapshot, the audit, the Repair Sprint, both monthly plans and the flagship each carry a live electric charge around their edge in their own colour: blue, green, amber, violet, cyan and flame. Fixed: the flagship label on the pricing page was unreadable (dark on dark); a large gap between the six dimensions and the report section on the audit page; the home-page hero form on phones is now two full-width pills instead of a button loose inside a tall box; and the magnetic buttons were borrowing the hero light’s coordinates on pointer devices, which shifted the hero’s primary button out of its row.

    SiteDesign
17. 14 Sep 2026

    Inner pages rebuilt with a shared kit: statement bands, icon cards, drawn rails, split features

    The audit, process, about, AI visibility, transformation, tools and sectors pages now share the rhythm of the home page. Each carries a dark statement band whose figures count up, feature cards with icons and a pointer spotlight, numbered rails that draw in as you scroll, and copy set beside a coded illustration: the report fanning out on the audit page, an assistant’s read-check-answer pipeline on AI visibility, week 0 against week 8 on the Transformation. The about page has a proper founder section, a pull quote and the clinic’s published figures. Also fixed: section titles and eyebrows were never revealing on real scrolling (a clipped box is never counted as in view, and titles inside a reveal block were never observed), which left blank gaps on the home page.

    SiteDesign
18. 14 Sep 2026

    Motion: a measures ticker, text that arrives in different ways, and animated illustrations

    The home hero now ends with a slow ticker of everything HighRegard measures. Headlines rise word by word, section titles sharpen from a blur, eyebrows wipe in, and one line on the Insights page types itself. Every hero and section introduction now starts faded and lights up word by word as it scrolls into view. Inner pages gained coded, animated illustrations beside the hero copy: a page being scanned on the audit page, a live prompt on AI visibility, a geo-grid filling in on the tools index, a rail of steps on the process page, a drawn chart on case studies, a ledger on pricing, a radar on about. The most useful pricing tier carries a live green charge around its edge. Story steps highlight as you scroll; the changelog rail draws in; primary buttons lean towards the pointer. On phones, the hero light now sits in the gutter instead of across the copy, proof points stack left-aligned, stacked buttons share one width, wrapped labels keep their marker on the first line, and the monthly-plan cards no longer overflow the screen. Everything respects reduced-motion and nothing is an image.

    SiteDesign
19. 13 Sep 2026

    Numbers the model cannot touch, and a recommendation QA layer

    The overview assessment now receives named, pre-formatted values and may only reference them; the application inserts every figure and rejects any text that adds, scales or re-units a number. The opportunity feed gained materiality thresholds, duplicate merging, a gate that blocks budget and bidding actions while tracking is broken, bidding-strategy-aware Google Ads advice, typed impact totals (pounds, conversions and clicks are no longer added together) and an observed / estimated / judgement split on every item. Data quality shows “not yet evaluated” instead of “0 failing”. Data sources now read Available, Coming next, Requires provider approval or Planned, with Maps, AI visibility, reviews and authority ahead of social sources. A /platform/ page shows the application’s real output; pricing states which purchases include a workspace; the privacy notice, cookie policy and data-processing page cover the application. Three new Insights pieces on the engine’s reasoning.

    ApplicationSiteLegalInsights
20. 13 Sep 2026

    Command palette and the new scene

    Press ⌘K (or /) anywhere to search every page, tool and action. The home page gained the light-beam scene, ghost previews of the instruments, and a cursor spotlight across the product tiles. Inner pages carry a lighter version of the same scene.

    SiteDesign
21. 12 Sep 2026

    Design system v2

    A two-scope token system (light body, black product surfaces), self-hosted Inter Tight and Inter, pill navigation with sign-in, and a home page that leads with the application rather than a description of it.

    SiteDesign
22. 10 Sep 2026

    The HighRegard application: opportunities, not dashboards

    Connect Search Console, Analytics and Google Ads read-only. Deterministic scoring turns them into a ranked queue of opportunities, each with a why, an expected effect and a verification plan. Tracking-health verdicts flag conversion signals that have stopped before you cut spend on a false zero. A live demo runs on synthetic data with no sign-in.

    Application
23. 6 Sep 2026

    Premises classifier and the snapshot workbench

    Free snapshots now classify whether a business is visited or remote before scoring local signals, so a service-area business is not penalised for having no public address. Internally, every snapshot lead opens a workbench with the evidence already gathered.

    ToolsMethodology
24. 4 Sep 2026

    Registered address and legal documents v1.2

    Contact, footer, terms, privacy and structured data now carry the Bradford address. The legal documents were re-issued as v1.2 with the same change.

    SiteLegal
25. 2 Sep 2026

    Product events and two new insights

    Tool starts, completions and checkout clicks are measured (consent-gated) so the free tools can be judged on outcomes. Two insight articles published, and the RSS feed now follows the article registry.

    SiteInsights
26. 1 Sep 2026

    Author page and the daily performance chart

    A founder author page with professional registration, a proof strip under the hero, and the case study gained a daily clicks chart rendered from the exported Search Console data, with the file fingerprint published.

    SiteCase study
27. 31 Aug 2026

    Geo-grid rank tracker calibrated

    Our own Maps geo-grid tracker was calibrated against manual checks across two towns. Scan cost is a fraction of a penny per point, which is what makes street-level Maps visibility affordable to publish.

    ToolsMethodology
28. 28 Aug 2026

    Live AI Visibility Test

    Three real customer prompts run against a live model with web search, returning whether you are named, who is named instead, and which sources were cited. Rate-limited and free.

    Tools
29. 25 Aug 2026

    Twenty free tools, one scoring methodology

    The Website Visibility Checker, AI Readiness Checker, Domain Authority Checker, Schema Checker and sixteen more, all scored against a published, versioned methodology with integrity hashes on the case-study exports.

    ToolsMethodology
30. 23 Aug 2026

    Asset versioning

    A deploy left returning visitors running day-old scripts against new pages. Every asset now carries a build hash, so the whole graph updates together.

    SiteFix
31. 14 Aug 2026

    First published case study

    Pulse & Clarity Clinic, Bradford: measured weekly in Search Console since 24 April 2026. Founder project, declared as such, with the export, the method and the limitations published alongside the numbers.

    Case study
32. 11 Aug 2026

    Launch

    HighRegard launches with one product line: the free Website Revenue Snapshot, the £295 Website Revenue Audit and the £795 Repair Sprint. Fixed prices, published scope, no sales call.

    Site