Drift alerts

The "Open ID Connect (OIDC)" section is now a word-for-word copy of the SAML section directly above it: same sign-in URL example (https://samlp.example.com/login), same X.509 signing certificate row, same Entity ID (urn:auth0:wellsaidlabs:XXX-saml), same ACS URL ending in -saml, and the same line "firstname, lastname, email — must be included in the SAML response." An Enterprise IT team setting up OIDC has nothing correct to work from, and every value on the page points them at a SAML connection instead. The four OIDC-specific values (Issuer URL, Client ID, Redirect URL, Initiative login URL) were published on this page as recently as the July 26 scan and disappeared in the July 31 rewrite.

The "Studio Update: A Simpler Way to Create Audio" entry links the words "WellSaid Studio" to https://help.wellsaid.io/hc/en-us/articles/48785974467731. That article, "Introducing the New WellSaid Studio", is still absent from the Help Center; requesting it returns "RecordNotFound — Not found", while a live article requested the same way returns normally. Nine articles were removed from the Help Center in the Aug 5–7 rewrite, and this is the one that still has an inbound link from a page customers use to catch up on what changed.

The July 30, 2026 transition date has passed. Both articles still read as advance notice: this one asks "What happens if you don't migrate by July 30," and Legacy Studio Migration FAQs says "You can keep editing and rendering in legacy projects normally up until July 30" and "If a project does become view-only on July 30." A customer opening a locked project today can't tell from either article whether the lock is expected or a fault. This is the same pattern as the existing finding on the June 1 migration launch article (51253958068755), but a different date on two different pages.

Building Your Team Voices List, republished this week, says "Click the bookmark icon on the voice card to add it to Team Voices" in both of its paths, and Favoriting a Voice spells out the difference: "Click the heart icon on the voice card to add it to Favorites" followed by "Note: Each voice card also has a bookmark icon next to the heart. That one adds the voice to Team Voices, a shared list for your team — it's separate from your personal Favorites." A new admin following this onboarding guide clicks the heart, and the voice lands in their own Favorites rather than the shared team list they were setting up. This page also still says "Choose a voice and style you wish to favorite from the Explore or All tabs" — the All tab was dropped from both of the rewritten voice articles.

This article was republished on 2026-08-07 and still answers "I'm running low on downloaded minutes" with "Add-on minute packs are also planned for a future release. Contact Support if you need additional minutes in the interim." Add-ons have been live for a month: the July 2026 Release Notes entry reads "Add-ons let you add extra minutes to your current plan – active the moment you buy", the article Adding More Minutes to Your Plan walks through buying one from Account Settings, Understanding Your Download Limit already points readers there, and every plan card on the pricing page now says "Extra minutes available". Customers who hit their limit are being sent to Support for something they can buy themselves in three clicks.

Article 51253958068755 (last updated 2026-06-05) still reads 'Starting June 1, 2026, you'll be able to migrate legacy projects', 'Migration is not automatic. Starting in June, you'll be able to migrate projects one at a time', and 'We'll be sharing a complete migration guide in June.' The stated date has already passed. Migration is already live: the companion guide 51751205121299 (published 2026-05-15) uses present tense ('you can migrate your legacy projects') and the dated timeline table lists 'June 1, 2026 — Migration opens.' Readers are told migration is a future capability and that a complete guide is still forthcoming, when both already exist. Same future-tense staleness pattern as the carried-forward May 20 findings (Sonny G. / Tobin A.) but on a different page and date, so not a duplicate of any carry-forward.

This page (https://www.wellsaid.io/features/features-security; the prior /features/security page was removed). Its Accessibility section reads 'We are committed to meet WCAG 2.1 and other national standards, working towards compliance in these areas by 2025.' The stated date has already passed. The rest of the section is forward-looking ('working towards'), so the page implies the WCAG 2.1 goal is still in progress past its own deadline. Not present in the carried-forward finding set.

This page contradicts itself: the "Pick a voice" step says "Access to 120+ voices, with a variety of accents, languages and tones" and the language section says "support for 30+ languages and 120+ studio-quality voices", while the voice showcase further down the same page says "Explore 280+ voices". Across the rest of the site: 280+ appears on eighteen pages including the homepage ("Choose from 280+ of the most realistic voices") and every use-case page; 240+ appears twice on /features/ai-voices ("Create studio-quality audio with 240+ realistic AI voices"); 120 appears on /solutions/advertising and in two blog posts. A buyer comparing vendors sees the library more than double depending on which page they land on. This finding was first raised in the 2026-06-28 report and the homepage half of it has since been fixed — the homepage now says 280+ in both places — but the five pages listed above still carry the older figures.

This page went live this week. The sentence above the plan comparison says "Paid plans add more minutes, commercial rights, and advanced features" — but the comparison directly below lists the free tier as "Trial $0/mo, No credit card required, 3 min/mo downloads, 10 min of generation, 3 projects, Commercial rights, Caption files". The pricing page describes the same tier as "3 download minutes per month (no commercial rights)". Every other row on the free column matches the pricing page (3 projects, 10 minutes of generation), so commercial rights is the one line out of step. Someone signing up for the free plan from this page could reasonably believe they can publish the audio commercially.

The August 2026 release note says "Access it anytime with the Generate All button in Studio." The article it announces, Generating Audio for All Sections, calls the same control Generate every time it names it: "Click on Generate in the top-right corner of the Studio", "the counter next to the Generate button counts down as each section finishes", and "If the counter still shows sections left, click Generate again to retry them." That article's wording is unchanged this week, so the two names still disagree. Someone who reads the release note and then goes looking for a Generate All button won't find one under that name.

Both articles were rewritten on 2026-08-05/07 and both added a plan requirement that the pricing page contradicts. This article now opens with "You'll need a Starter plan or higher", but on wellsaid.io/ai-voice-pricing the Starter card lists no Adobe integration at all, the Pro card lists "Adobe Express integration", and the comparison table's Adobe integrations row reads "Adobe Express integration | Adobe Premiere Pro & Express integrations | Adobe Premiere Pro & Express integrations" across Pro, Business and Enterprise only. The new /free-ai-voice-generator page agrees, showing Adobe Express integration on Pro. Separately, Collaborator Role now leads with "Collaborator licenses are available as an add-on to Enterprise subscriptions", while the pricing page's Business card reads "Everything in Pro, plus: Purchase up to 5 seats, Team workspace, Collaborator seats, Live chat support". A Starter subscriber installs the Adobe add-on and finds it unavailable; a Business customer is told to buy Enterprise for something their plan already lists.

The intro paragraph added to this article this week says the list covers voices "each tagged with the style it supports — Narration, Conversational, Promo, or Character". Of the 205 voices listed, 109 are tagged "(Partner)" instead — every Arabic, Chinese, Czech, Danish, Dutch, Italian, Japanese, Korean, Persian, Polish, Portuguese, Swedish and Turkish voice, plus most Spanish, French and German ones. Partner is a voice model, not a style: Creating Audio at Different Sample Rates says "Standard and Partner voices always render at 24 kHz", and Exploring Languages, Accents, and Styles says "Voices are organized into four speaking styles" with no Partner among them. Someone looking for a Conversational Spanish voice can't tell from this list which Spanish voices offer one.

The question is "Can I apply more than one preset to a section?" but the answer talks about applying presets across multiple sections — a different thing. Read literally, it also contradicts the rest of the page: the Removing a preset steps describe a single Reset icon that returns the voice to its default, which only makes sense with one preset per section. Until the July 31 rewrite this FAQ answered "No. Only one Emotional Preset can be active per section at a time. Applying a new preset replaces the previous one."

The "Supported languages & accents" list appears on fifteen marketing pages, including this one and the homepage, and every copy of it lists English (New Zealand). The Help Center's List of WellSaid Voices — which is organised by accent and names every voice — has no New Zealand section at all. The only New Zealand label anywhere on the site is a single voice card on /text-to-speech-api reading "Issa B. English (NZ)", and the homepage, /features/ai-voices, /ai-voice-use-cases/video-production and /free-ai-voice-generator all label that same voice "Issa B. English (ZA)", matching the Help Center's English (South Africa) section. In the other direction, the list omits English (Mexico) even though this page's own hero plays "Hannah A. English (MX)" and the Help Center has an English (Mexico) section with three voices, and it omits English (United Kingdom) despite UK voices being demonstrated on this page ("James B. English (UK)") and across the site. Someone shortlisting vendors for a New Zealand audience would choose WellSaid and find no voice to use, and someone who needs a UK or Mexican English voice may rule WellSaid out for lacking one it has.

The pricing page's Export formats row now reads "MP3MP3MP3, TXTMP3, WAV, OGG, TXTMP3, WAV, OGG, TXT" across Trial, Starter, Pro, Business and Enterprise — the Pro column changed from "MP3, WAV, OGG, TXT" to "MP3, TXT" this week. This guide, rewritten on 2026-07-25, still says "Choose a file format: .mp3, .wav, or .ogg." with no plan condition, and Downloading Sections and Projects says "OGG and WAV are available on Business and Enterprise plans; MP3 is available on all plans." A Pro subscriber following the getting-started steps will look for a .wav option that their plan no longer offers.

This page was updated this week. Its "Multilingual employee comms" tile reads "Deliver the same message clearly across 17 languages and 280+ AI voices without rebuilding anything", while the "Supported languages & accents" table on the same page lists 18 languages. Two other figures are in circulation: 30+ (on /industries/voice-ai-marketing and in the Help Center's List of WellSaid Voices) and 50+ ("Access 50+ languages including Spanish, French, Japanese, Korean, and more" on the homepage, "50+ languages & accents" on corporate-training and social-media-content). A communications lead checking whether their regional languages are covered gets 17, 18, 30+ or 50+ depending on the page. The voice-count half of the same sentence is covered separately by 'marketing-voice-count-inconsistency-120-240-280'.

Two FAQs above this one on the same page deal with 96 kHz files: "Files rendered at 96 kHz are limited by total file size" and "The 96 kHz files will be downsampled in the combined output." The pricing page lists Enterprise sample rates as "24 kHz, 48 kHz, 96 kHz." An Enterprise customer reading this FAQ could reasonably conclude 96 kHz isn't offered. This line carried over unchanged when the article was renamed from Using the Combine Tool on August 1.

This article was rewritten again this week. Its Choosing a caption format section lists exactly two — "SRT (SubRip Subtitle) – works with most video players and editing tools" and "VTT (WebVTT) – built for web use" — but the download step below says "Select a caption format: SRT, VTT, or TXT." The product itself separates them: Updating Global Download Settings tells readers to "enable .txt files if needed" under audio format, then separately "Select the checkbox for SRT or VTT (or both)" under Include closed caption file. The pricing page's comparison table also lists "Caption file downloads .SRT, .VTT" for every paid plan. A customer checking what caption formats their plan includes gets a different answer than the download step gives them.

This article was rewritten this week. It defines Can View as "Enables viewing and listening to voiceovers without editing capabilities plus downloads. Ideal for team members who only need access to review projects" — then, two paragraphs later, tells readers that "Anyone with Full or View Edit Access can remove a team member's access to a project entirely, including the project owner." Read as written, a review-only member can lock the project owner out of their own work. It also invents two permission names the article never defines: the personal-file section says "Anyone with Full Edit Access", the Workspace section says "Full or View Edit Access", while the permission picker in both sections offers Full Access, Can Edit, and Can View. An admin setting up review access for stakeholders can't tell from this page what a Can View member is actually able to do.

This Help Center article (updated June 29) says 'Enjoy a 14-day trial period to explore and test our API capabilities.' The public API page at wellsaid.io/text-to-speech-api says 'Signing up for an API trial is free and easy. You'll have 7 days to try out all of our features.' Both describe the same API trial, so a developer comparing the two can't tell whether it's 7 or 14 days.

This article was rewritten again this week (1,167 to 584 words, updated 2026-07-28) and still mixes the two names. It says "Respellings are created inside the Replacements panel" and "Click on + New replacement in the Smart Toolbar", then "Saved respellings go to your Pronunciation Library", "Save the entry to your Pronunciation Library" and "re-save it in the Pronunciation Library". The Guide to Replacements, updated the same week, tells readers to "Open Libraries in the left navigation" — so two freshly-updated articles name the same place differently. A reader following these steps looks for a "Pronunciation Library" that isn't labelled that way in the product.

This Help Center article, edited on 2026-09-08, lists chat as available for Business and Enterprise customers and, under "Response and resolution times", states "Chat: Within 1 hour during open hours." The pricing page (https://www.wellsaid.io/ai-voice-pricing) instead promises Business "Email (1 business day SLA) and Live Chat (4 hour SLA during business hours)" and reserves the 1-hour figure for Enterprise: "Priority Chat (1 hour SLA during business hours)". A Business customer reading the support article expects a 1-hour chat reply that the pricing SLA doesn't guarantee.

This paragraph was added to the article this week. The pricing page describes the free tier as "3 download minutes per month (no commercial rights)" plus "Curated voice actors", and reserves "All English voices" for the Starter and Pro cards. So the one tier in the stated range that doesn't get the full English library is the one the sentence starts from. A prospect on the free plan reads this article, goes looking for a specific accent or style listed below it, and doesn't find it.

The use-case cards on the homepage were rewritten in the latest site update (the card descriptions now match their headings, which fixed an earlier mix-up). The new "Podcasting & radio" card copy reads "Keep narration, consistent across episodes, segments, and recurring audio content." — the comma after "narration" splits the phrase "keep narration consistent" in half. Nearby, the "Social media content" card now writes "reels" and "stories" in lowercase while keeping "TikToks" capitalized; worth aligning capitalization in the same edit if those refer to Instagram Reels and Stories.

Both pages introduce the integrations with "Try our integrations — Adobe Express, Adobe Premier Pro", dropping the final e from Premiere. The same Marketing page spells it "Adobe Premiere Pro" three times in its own body copy, and so do the pricing page and the Help Center guide at help.wellsaid.io/…/39177295482387-WellSaid-Voiceover-for-Adobe-Premiere-Pro. It's a small thing, but it sits in the integration callout a marketer is most likely to search for by name.

Corroborated by existing WellSaid Mode B audit: 'Article says See All Avatars -- live UI says see all voices'. Current-app capture confirms no 'Avatar' terminology anywhere in the voice picker or right rail.

Mode B audit (2026-04) and current-app capture both confirm Studio no longer uses 'avatar' anywhere in visible UI. The voice-variant combobox's aria-label still reads 'Select avatar variant', but that text is only read by screen readers — sighted users never see 'avatar'.

Current-app capture: 'New project' modal shows three radio cards -- 'WellSaid Studio' (RECOMMENDED), 'Legacy Studio - Standard', 'Legacy Studio - Global' (with additional language dropdown).

Corroborated by an existing WellSaid finding: 'Article says Voice Library — the dialog is now titled Voices'.

Current-app capture of the projects home shows '+ New' button in the top-right; a second capture shows the popover menu with Project/Folder/Workspace options.

Current-app capture of the New project modal.

Current-app capture of the editor right rail.