What's New

Recent product improvements, newest first. Dates are when the change reached everyone.

November 2026

• Posting a job from a candidate account makes you a recruiter. If you already have a candidate account and post a job through the e-mail link, your account is upgraded to recruiter instead of being left as a candidate; a paid individual plan is told why it cannot be converted and offered a switch-account path.
• Your resume picks the suggested role. On the interview setup page the suggested role now comes from the title on your resume. If the resume has no clear title, the field stays empty for you to choose, and a resume uploaded in the Resume Optimizer becomes the selected resume on the setup page.
• Start is always in view on interview setup. The Start button stays on screen while you scroll the setup page, and the mode tooltip now explains how long each round takes.
• Entry-level coding rounds are entry-level. Technical rounds with coding at the entry level now ask questions one step easier than before, matched to the level you chose.
• A recording interruption pauses the interview. When an interview must be recorded and the recording stops, the interview pauses and "Retry Recording" resumes it instead of losing the session.
• Integrity flags no longer end an interview by default. A flagged signal is recorded for the recruiter's report; ending the interview on flags is now a platform-wide setting that is off.
• The login page's Candidate tab always opens sign-in. Switching between Recruiter and Candidate on the login page no longer shows a stale create-account view.
• Invited candidates go straight in. If the role you were invited to does not require identity verification, clicking Start now takes you directly into the interview, with no government-ID step on the way.
• Integrity flags name each signal. When two integrity signals happen together, both are recorded instead of only the first, and switching away from the interview tab now shows up as its own flag.
• Integrity details tucked behind one chip. On the feedback report, the integrity section for recruiters starts collapsed behind a single summary chip; open it to see every signal.
• Redesigned feedback and interview setup pages. The interview feedback report and the pre-interview setup page have a cleaner layout that is easier to read on a phone.
• Recording really starts before your interview does. When an interview must be recorded, the room now waits until recording has actually started, retries once if the first attempt stalls, and asks you to confirm before you leave.
• Resume Optimizer spots duplicates. Uploading a resume you already have tells you so, and the page shows how many resumes your plan stores.
• Profile Card owners see share counts. Your Profile Card page shows how many times you shared it.
• Privacy request form. Ask for a copy, correction or deletion of your data at theinterviews.ai/privacy/data-request.
• E-mail from hello@ and notify@. Messages from TheInterviews now come from hello@theinterviews.ai (people) or notify@theinterviews.ai (automatic notices); replies go to hello@.
• Brand lockup on share cards. Link previews and the public Profile Card page carry the final TheInterviews lockup.

October 2026

• Watch the TheInterviews film. The homepage now has a short film showing a practice interview, the report and how hiring teams use it. Press "Watch the film" at the top of the page, or share theinterviews.ai/watch/home.
• Interview recordings stay private to the hiring team. Recruiters now play a candidate's recording, captions and question markers only while signed in to the posting that owns it. There is no shareable storage link any more.
• A new logo across the site and apps. The site, browser tab and app icons use the final TheInterviews mark.
• One contact address. Reach us at hello@theinterviews.ai. The contact page also names the company, TheInterviews Inc.
• Smoother practice setup in the mobile app. The app no longer pre-selects a round your phone can't start, and the setup page explains why Start is unavailable when it is.
• Pay for your first plan by US bank account. When you buy your first paid plan, checkout now offers US bank debit (ACH) alongside card and wallets.
• Deleting your account also stops billing. Deleting your account now cancels any paid subscription and emails you a confirmation. You can also start a deletion request from theinterviews.ai/legal/delete-account.
• Report an AI feedback response. If something in your feedback report looks wrong or inappropriate, use "Report this AI feedback" under the report to tell us.
• Pages load faster. Shared interface code now downloads once and is reused across pages, so the site loads about 100 KB less in total.
• See who is looking at your Profile Card. Your Profile Card page now shows how many times the card was viewed, how many share links you created and how many times those links were opened.
• Sharing your interview link shows a preview card. When you paste your candidate share link into LinkedIn or a messenger, it now shows the role, the organisation and an image.
• Job posting candidate counts are right. A posting's dashboard shows how many candidates have interviewed and how many seats remain, and a candidate who retries doesn't take a second seat.
• Human star ratings land on the right scale. A 1–5 star rating from an interviewer is now placed on the 0–100 Profile Card scale instead of being used as a raw number.
• Interviewers can mark a no-show. Ten minutes after the start time, if the candidate hasn't arrived, the interviewer can record a no-show from their upcoming-interviews page, which also shows which sessions come from our interviewer bench and the state of each report.
• You won't be asked the same question twice. Practice interviews now remember what you have already been asked in earlier sessions and pick something new.
• Skills you didn't get to show are marked "Not assessed" instead of a number, on your feedback report and your Profile Card, and they no longer pull your average down.
• Declined answers are marked as not answered. If you say you can't answer a question right now, it shows the "Not answered" badge again.
• The Free plan limit message is correct again. When you reach your Free practice allowance you now see "Practice interview limit reached" rather than "Your subscription has ended".
• Upgrade suggestion for Lifelike practice on the Free plan. If you have taken Lifelike interviews on a Free plan, your dashboard now suggests upgrading to keep taking them.
• Evidence behind your Profile Card scores. You and recruiters you share with can expand each skill to see short excerpts from your interview, and you choose whether they are shown.
• Recruiter feedback on reports. Recruiters can mark each skill in a report as fair or not and leave a note; this doesn't change your score.
• Export your verified credential as a badge. Download it from your Profile Card as an SVG or PNG image to use anywhere you like.
• Feedback rewards real examples. Behavioural answers are now judged on the story you tell (what happened, what you did and what came of it) rather than on whether you followed a named framework.
• Avatar interviews start more smoothly. The interviewer waits until the avatar has joined before greeting you, avatar audio problems are detected and fixed without restarting the avatar, and a dropped connection while the avatar is warming up no longer counts against your interview.
• Upload Resume on your profile banner works in place and fills in your profile, without leaving the page.
• Coding scores appear only when you wrote code, so a spoken-only interview isn't marked on a skill it never tested.
• Referral bonus interviews are simpler. They apply to practice interviews, are used when the interview starts, and no longer count against your plan's allowance.
• Joining an organization keeps your existing shares. Your shared feedback links stay valid when you join.
• First-time visitors are greeted with "Welcome!" rather than "Welcome Back!".
• Faster pages on phones. The homepage and the sign-in page load sooner, and the Pricing page no longer jumps as it loads.
• Try the demo on a specific question. A demo link can now open the demo interview on one chosen question, and the homepage shows how a recruiter sees a Profile Card.
• Only you or our support team can deactivate your account.
• Practice interviews ask follow-up questions again. When you choose AI-generated questions, the interviewer now stays in that mode for the whole session and follows up on what you actually said.
• Your Role and Interview Type picks survive a résumé upload. A résumé's recommendation is offered as a suggestion you can apply, instead of replacing what you chose.
• The feedback page tells you when your report is still being written, and fills in on its own when it is ready. Every failure state now offers "Try again".
• The post-interview survey keeps what you typed through a refresh, a failed submit or closing the dialog, and once skipped it stays skipped.
• Behavioural reports no longer show a coding section when you wrote no code, and every report now lists your top gaps alongside your strengths.
• Saying "I don't know" is no longer graded as an answer.
• The coding round shows the interviewer's captions, like every other round.
• Long feedback PDFs paginate instead of stopping after one page.
• A welcome email now arrives when you sign up with email and password, not only with Google.
• Pricing is easier to read. One clear button per plan, and a zero price simply says "Free".
• Redesigned Network Reach page and alignment fixes on the interview setup screen and the Manage Subscription button.
• Role and Interview Type choices stay available on the AI interview setup screen. They no longer come up empty after you reload the page or return from an interview.
• Resume Optimizer now needs a sign-in and has a monthly allowance. You get 3 runs on the Free plan and 25 on paid plans, counted over a rolling 30 days, with a counter showing how many runs you have left.
• Importing a résumé during sign-up includes a quick human check. It takes a moment and keeps automated abuse out.
• Your name survives a page reload during sign-up. The first and last name you typed are kept if the page refreshes.
• Redesigned Notifications, Shared Feedback and My Profile Card pages, plus a verified-interview badge on your Profile Card.
• Interview setup fixes. Several small corrections to the setup flow before you start.
• Time-boxed AI interviews end the same way every time. Running out of time now leads to a single, consistent outcome.
• Recordings are no longer lost when a file is slow to appear. We now retry instead of dropping the recording.

September 2026

• A cleaner top navigation. The header no longer repeats a link to the homepage; About and Contact replace the old labels, and a few small spacing fixes tidy up the Pricing menu.
• A brief connection hiccup no longer signs you out. If the app loses touch with our servers for a moment, you now see a short reconnecting banner and keep your place, instead of being logged out.
• Your practice quota is described honestly. The free-plan practice screen now tells you plainly when your next free interview opens, in your own local time.
• The post-interview survey asks for less. Only a star rating is required; the written comment is optional.
• A quick device check before you join. Before entering the interview room you now land in a short lobby that checks your camera and microphone are working.
• Coding rounds are easier to navigate by keyboard. Pressing Tab after leaving the code editor now reliably moves you to the next control.
• Résumé matching reads what is actually on the page. A résumé's content, not its filename, is what now grounds AI-based matching and feedback.
• /practice takes you somewhere useful. The old placeholder page now sends you straight to candidate sign-up.
• Interviewers can sign in again. Anyone with an interviewer account was bounced between the home page and the dashboard after logging in, and then signed out with a "Connection Problem" message. Signing in now lands on the dashboard, and if a redirect ever loops again you get a page with Retry and Sign out instead of a spinning tab.
• Interviewers always see "Interview Others". The Practice → Interview Others entry, where an interviewer offers their availability, is shown to every interviewer regardless of plan or platform settings — it is their job, not a plan feature.
• Peer-to-Peer and Interview Others follow your plan. These entries appear when your plan includes peer interviews (every Allied tier does) or when the platform has them switched on for everyone; if neither applies, the entry is simply not shown and the availability page explains it is a members-only feature — no more buttons that fail after you click them.
• Professional Interviews only appear when your plan includes them. Members see the human mock-interview bench; everyone else sees nothing in the sidebar and a short members-only note on a direct link, instead of an error and a retry button.
• The Allied programme opens by invitation. Allied is a small, invitation-only tier for candidates who want a human career manager alongside the AI interviews. It is not listed on the pricing page: an invitation link from us shows the Allied options on the pricing grid and in sign-up for the person who tapped it, and nowhere else. Every Allied seat includes a written Services Agreement you read before you pay, and the number of seats is capped — when they are taken, the same link offers a waitlist instead.
• Plans with a fixed term now say so on the Billing page. A plan that does not renew on its own reads "Ends on <date> — renew to keep your seat. No automatic charge." with a Renew button; Cancel and Reactivate no longer appear on such a plan, because there is nothing to cancel.
• Human mock interviews included in a plan show a running count. On the professional interviews page a plan that includes human sessions shows "n of N left this month" and the reset date; a session covered by the plan is booked without a payment step and is labelled "Included · covered by your plan".
• Some question banks and resources belong to a tier. Where your plan does not include a bank or a document, it no longer appears in the list; if a higher tier has material you cannot see, one locked card says so and links to the application page rather than showing a price.
• Platform resources can only be uploaded by the content team. Any signed-in account could previously upload or remove a shared resource; that now needs the content-editor seat.
• The homepage Profile Card now shows seven candidates, not one. After the first card assembles, the hero cycles through a nurse, an IT support technician, a social worker, a teacher, a logistics coordinator and a pharma QC analyst; the pill for the card on screen fills across its turn, a pill jumps to that field, and a pause button stops the rotation. Reduce Motion gets the still card.
• A phone can no longer start a coding or system-design round. Those rounds need a keyboard and a wide screen, so the setup page on a phone shows a "use a laptop" card with a copy-link for the laptop instead of a Start button; voice rounds are unchanged.
• Coding questions open in your own language. The setup screen offers a "Coding language" choice (or Auto, which reads your résumé), the editor opens in that language with a matching starter function, and a Python or Java answer is no longer marked down for "wrong language".
• Everything on a phone is now a real tap target. Every button and control is at least 44 px, the top-bar menu button stays on-screen on narrow phones (it was pushed off the right edge), the hero buttons go full-width, and the accessibility statement now says what the interview room supports on a phone.
• The interview room loads when you start an interview, not when the page opens — the practice page downloads a third less JavaScript up front.
• Light-mode fixes on sign-up and the blog. Values typed into a sign-up field were white on white; several sign-up and blog labels were below the contrast threshold. All fixed.
• Clearing your phone number in the profile editor actually clears it, and a single-character first name is accepted.
• Search and social previews no longer advertise human interviewers or peer practice, and the Help Center's "schedule an interview" answer describes starting an AI interview.
• Candidates no longer see integrity details on their own report — only a short neutral note.
• Going Back to the sign-up page after verifying your e-mail keeps it verified.
• Rate limits now key on your real address. A client forging proxy headers could mint itself a fresh allowance on login, sign-up, contact forms and the demo; it cannot any more. Legitimate users see no change.
• A real favicon and web manifest — bookmarks, Slack unfurls and Android home screens show the brand mark.
• The AI interviewer no longer asks a question you have no time to answer. When the clock is nearly out, a round now wraps up instead of opening a fresh question, and an answer you were in the middle of giving is kept rather than lost when time runs out. A spoken command such as "next question" at the end of an answer is no longer transcribed into the answer itself.
• Your daily avatar allowance is now one budget across both avatar providers. It used to be counted in minutes for one provider only, so the other provider had no limit at all. If you use the allowance up, the interview continues with the local-video interviewer rather than stopping; the budget is set so that a normal day of practice never reaches it.
• Names in Devanagari, Tamil, Telugu, Kannada, Bengali, Gujarati and Malayalam are accepted at signup and on your profile, and typing one no longer loses its vowel marks as you go.
• Starting an interview when your allowance is used up now tells you why. Whether it is your plan's allowance or your organisation's ceiling, you see the real reason and, for a plan limit, a direct route to the subscription page. Every refusal had been collapsing into one generic message.
• Moving to a cheaper paid plan now takes effect straight away, with the unused part of your current period credited, instead of waiting for the period to end.
• An invitation is claimed the moment your account exists, however you signed up or signed in. Invitations had been staying pending when the invited person joined through a different route.
• Profile Card governance corrections. An approved card is no longer refused as "superseded" because publishing it changed its public/private setting; only the organisation that governs you can revoke your approved card; regenerating a card goes through the same approval check as publishing it; and the 60% profile-completeness rule now applies on every route that can publish.
• A momentary look away no longer ends a Lifelike interview. Termination now needs a sustained episode — several distinct no-face or multiple-person stretches — rather than three individual frames.
• .docx résumés keep their header blocks. Contact details placed in a document header are now read, and the résumé parser no longer depends on a component with a published vulnerability.
• The Privacy Policy and Terms name the company that operates the service — TheInterviews Inc., a Delaware corporation and the successor to Invent Artificial LLC — and describe the privacy regime that actually applies to candidate video.
• We have a new look. The mark, the wordmark and the type and colour scale across the site have been redrawn. Nothing about how the product works has changed, and the interview room keeps its own appearance.
• Organisations can now govern who publishes a Profile Card. A university or employer can require that a member's card is approved before it becomes visible. A member who needs approval is offered a way to ask, with a reason, rather than a button that does nothing; administrators get a queue, a preview of the card exactly as it would appear, and a note recorded against every decision. An approval is bound to the version of the card it was granted for, so it cannot be reused for a card that has since been rewritten. Revoking a card, or offboarding the person who owns it, stops every route to it at once — the published page, any share link already handed out, and its listing in Talent Discovery. A seat granted by an organisation also sets the member's plan while it lasts, so a student on the free tier gets the organisation's entitlements and returns to their own plan when the seat ends. Organisations that do not use this see no change.
• The interview room can be completed with a keyboard and with a screen reader. Every step of a live interview is now reachable without a mouse, and accessibility is checked automatically on every change rather than reviewed occasionally.
• A report no longer says a skill was "not measured" when your interview did measure it. Several skill names used by the technical, system-design and general interviews were not reaching the five dimensions on the report card. In an audit of recent reports, roughly seven in ten were dropping at least one. Scores that were already shown are unchanged.
• A long answer no longer ends your interview early. Working quietly through one long coding or system-design question used to risk the session being marked abandoned while you were still in the room. And if a question ever reached a dead end, the interview now moves on instead of leaving you waiting for something that was never coming.
• Releasing an update no longer interrupts interviews in progress. A server being updated stops accepting new interviews and waits for the ones already running to finish.
• The completion screen now tells you the outcome it recorded, rather than leaving you to infer whether the interview finished.
• Saving your profile no longer fails because of a field you did not edit. Every save re-checked your phone number and your name, whatever you had actually changed — so a value already on file that does not fit today's rules blocked the whole save. Most accounts have no phone number at all, because signing up never asks for one, and a smaller group hold a name that a later, stricter rule rejects. Either way the message was "Please fix the highlighted errors before saving" and nothing was saved, which also meant a headline or a location could not be set — and those count toward the profile completion a Profile Card needs before it can be published. Both checks now apply only to a field you have changed. The phone field no longer describes itself as required either, and clearing a phone number that cannot be cleared now says so, instead of reporting a save that did not happen.
• A rating you never earned is no longer shown. When an interview did not exercise a particular dimension, the report used to fill that axis with your overall score under the dimension's name. It now says the dimension was not measured — on the radar chart, in the written report and in the PDF. On the sessions we replayed, about four axes in ten had been showing a number that was never measured. Nothing that was measured has changed.
• Recruiters now receive the "report ready" email. When a candidate completes an interview for one of your job postings, the email that tells you the report is ready is sent. It had never been sent before: interviews were not being linked back to the posting they were run under. Interviews completed before this release are not linked retroactively.
• The rubric your interview is scored against is now published in full. /scoring sets out every dimension an interview is graded on, the score bands, the number of interviews behind each figure with the date those counts were taken, and — in the same page rather than a footnote — the measured limits of the score. Nothing on it is illustrative; every number is attributable.
• We published our own evaluation of the scorer, including what it gets wrong. On a corpus that an independent rater separates into clearly distinct groups, 124 of 141 per-dimension scores came back as exactly 90 out of 100. That is the scorer compressing genuinely different performances into one number, and it is our finding about our own product. The method, the corpus, the questions we committed to in advance and the raw results are at /methodology/scorer-compression, published unchanged.
  note

  This is why /scoring states the limits of the score alongside the rubric. If you are reading a Profile Card rating, read both pages together.

• Five role pages that answer the question directly. /interview-questions now has written answers for software engineer, full-stack engineer, Python backend engineer, data engineer and AI/machine-learning engineer — drawn from real sessions, not invented for the page.
• The pricing page is now reachable. The footer and the site map both pointed at a section of the homepage rather than at /pricing, so the actual pricing page was linked from nowhere. That is fixed, and the same pass rebuilt the site's internal links around role hubs: pages you could not reach from the homepage went from 25 to 5, the site map no longer lists anything orphaned, and no page is more than three clicks from the front door.
• What a search engine or an assistant reads about a page now matches the page. Titles and the descriptions that appear under them in search results were re-cut to the length a result actually shows, and the structured data behind our marketing pages was extended and corrected so that every claim it makes is one the visible page also makes.
• Legacy hairlines are no longer white in dark mode. Table rules, blockquote rules, input outlines and empty rating stars were painting a bright white line on a dark background. They now follow the theme. A contrast check over our colour tokens also runs on every build, so this class of problem fails the build instead of reaching you.
• Pages that are not a live interview no longer download the interview room's styles. Six stylesheets that only the session room needs were removed from the bundle every page loads, cutting the site-wide stylesheet by about a fifth.
• The signup page stops moving under you as it loads. The candidate signup form used to mount after the navigation and footer, so the page jumped twice while you were reaching for it. The form now holds its space from the first paint.
• The upgrade prompt speaks to work you have actually done. For accounts on the free plan the dashboard prompt used to describe the paid plan in the abstract; it now refers to the practice interviews you have run. Where we cannot evidence that count, it says nothing rather than guessing.
• We now record how an account arrived. If you reach us from an AI assistant, a search or a campaign link, the account records that source — including when the assistant strips the usual referrer, and now on Google sign-up as well, which previously recorded nothing. It stores where the visit came from, not anything further about you, and it exists so we can tell which of these pages are worth writing.

August 2026

• The homepage header is now just our name. The sub-line under the wordmark made a claim about a single verified standard that the product does not yet keep, so it has been retired — from the header, both footers, the customer navigation, share cards and the site's structured data. The page title and the description that appears under it in search results have also been resized: the description had been more than twice the length a search result will show, so most of it was never read by anyone.

• The pricing page can be read without JavaScript. Plans, prices and currency on /pricing are now produced on the server, so they are present the moment the page arrives — including for assistants and search engines that do not run scripts.

• We now state in writing which assistants may read the site. /llms.txt and /llms-full.txt set out plainly which AI assistants are welcome to read TheInterviews.ai and what they may read.

• The setup form no longer offers a length that coding rounds ignore. Choosing a coding interview showed a length picker, but a coding round does not use it — so the form promised a duration the interview would not keep. The picker is now hidden for those rounds, and the interview runs exactly as long as the form says everywhere it is still shown.

• An interview that ends early now records why. When a session stops without finishing, the reason is kept — leaving deliberately is told apart from simply closing the tab — along with how far the interview got and how long it actually ran. Nothing about this is visible during an interview; it exists so that a session that goes wrong can be understood afterwards instead of disappearing silently.

• A link you share now shows a preview card of its own. Sharing a marketing link such as theinterviews.ai/free into WhatsApp, LinkedIn or Slack used to hand the preview to whichever page the link pointed at, so the link you actually sent had no card. It now carries its own — showing the same title, description and image as its destination — and the preview keeps the link's own address, so a colleague who re-shares it still credits whoever's link it was. Tapping it takes you to exactly the same page as before, and Back returns you to where you came from rather than to the link.

• Interview pages start with less to download. The library that powers a live session is now fetched only when a session actually begins, instead of on first paint of every page that might need one.

• Your Profile Card rating now counts everything the interviewer assessed. After each AI interview the interviewer scores you on several named skills. Depending on which interview you sat, some of those scores — and for some interviews all of them — were being discarded before they reached the rating on your Profile Card. Nearly a quarter were lost overall. The one most often lost was the assessment of your hands-on coding, which is why Code Quality showed a real measurement on very few interviews despite being a substantial part of a software engineer's rating; it now shows one on more than half. Where an interview's scores were all discarded, the whole interview was replaced by a single overall number standing in for every skill area.

• Ratings that had silently stopped updating are moving again. For some accounts the rating had frozen: every attempt to refresh it failed quietly, so completing more interviews changed nothing. Those accounts are now recalculated.

  note

  Both fixes above are corrections, so some published Profile Card ratings have changed — in either direction. Every rating has been recalculated from your actual interview results under the current scoring rules, and where an account has no assessment that counts toward the rating, it now shows no rating at all rather than a zero.

• On a phone, the Start button is back where you expect it. On the AI Interview Simulator setup screen the "Start AI Interview Simulator" button sat at the very bottom of the page, below the video preview and readiness panels — so you would finish the form and find no button there. It now sits directly under the form you have just filled in. Two things went with it: a long resume filename no longer stretches the whole page sideways on a phone, and the interview-mode dropdown is legible again in the light theme.

• Your feedback report has been redesigned. The summary, skills gap, your next steps, JD alignment, the MAANG+ bar, the skills radar and the quality judge have all been reworked for the light theme. Several rough edges went with the redesign, including a leftover developer note that could appear in place of your strengths, a "code replay isn't available" panel that showed on reports that never had code, and a score that could read as "NaN%".

• Team rosters apply the roles you set. If you administer an organisation, the member roster template you download now matches what our importer reads. Previously it did not: a row you marked as an administrator was quietly imported as an ordinary member and the upload still reported success, so you could not give a colleague administrative access and nothing said why. Cohorts are separated with a semicolon, so listing two of them no longer creates one oddly-named group, and a file missing the role column is now reported row by row instead of being treated as though every role had been left blank.

• A burst of traffic no longer takes the site down. On 21 August a single source sent requests fast enough to occupy every available database connection, and for about four minutes the site returned errors — to people signed in and to visitors who were only reading the homepage. Requests that can't get a connection are now turned away in seconds rather than queuing for thirty, so a burst is absorbed and the site recovers on its own instead of falling over. We can also now adjust those limits on a running service within thirty seconds, which we were unable to do while it was happening.

• Role suggestions now cover your whole specialty. Upload a resume and you'll see the related roles in your field, not just the generic titles. Someone working on large language models is now offered NLP, Computer Vision and MLOps Engineer; a Kubernetes engineer gets Terraform and Cloud Engineer. Previously 85 of our 217 roles could only appear if your resume headline named them exactly.

• Roles for work beyond software. The role list now covers recruiting and staffing (including bench sales and US IT recruiting), HR, sales, finance, operations, legal, education, healthcare, marketing and traditional engineering — so uploading a resume suggests roles from your own field instead of a list of software titles.

• Setup suggests a role from your resume. Upload a resume at setup and the role list narrows to your field, with a suggested role pre-filled and interview types ordered to suit it. If we can't tell, you still get the full list rather than an empty one.

• A refreshed look across your account. The dashboard, your profile, the AI Interview Simulator and the subscription & plans pages have all been redesigned — same features, laid out more clearly.

• Pausing an interview works properly again. Putting an interview on hold and coming back now resumes your thinking time where it left off, stops cleanly once that time is used up, and no longer echoes audio when you return.

• The role picker works without a mouse. The role selector on the interview setup screen can now be operated entirely from the keyboard and is announced correctly by screen readers.

• Recruiter session recordings are back. Recordings of recruiter-led interview sessions save reliably again after a fix to our recording pipeline.

• Find us on X and YouTube. The site footer now links to our official X (@theinterviewsai) and YouTube channels.

• Buttons take you where you already are. If you're signed in, marketing "Get Started" buttons now take you straight into the product instead of back to the Sign Up form.

• Plan pages read cleanly. Every plan card now shows its full description (a display bug was cutting the last line of text in half), and upgrade prompts always match the plan you're on.

• Cancellation emails arrive exactly once. Cancelling a subscription from your card provider's side now sends a single confirmation email, and renewal payments are always recorded against your order correctly.

• Planned maintenance now looks planned. When we schedule maintenance you'll see a banner with the window ahead of time, and in the rare case the site is unavailable you'll get a clear "back soon" page instead of an error.

• Smoother releases. We changed how updates roll out so the site no longer serves a mix of old and new versions during a release.

• Refunds show up correctly. A refund on a subscription purchase is now reflected on your order immediately.

• The consent notice in the interview room matches exactly what we do — including the correct evidence-retention window (30 days).

• Your interview runs the full scheduled length. Sessions of every length (10–30 minutes) now keep the interviewer listening for the whole session.

• A calmer way to finish an answer. The countdown bar that pressured you while you paused mid-answer is gone; a quiet indicator shows the interview is simply waiting for you.

• Say "I'm done" to move on. A spoken command can submit your answer and move to the next question once you've actually said something substantive.

• Industry-specific questions. The role you choose at setup now drives industry-tailored question packs.

July 2026

• Take a moment to think. Say "give me a second" and the interviewer waits — with a calm indicator instead of a countdown — until you say you're ready (or press H).
• Fairer, comparable scoring. Graded (Lifelike) interviews all run the same length and difficulty, so recruiters compare candidates on one scale. Practice interviews keep the full choice of length and difficulty.
• Faster first question and a more reliable connection during answers.
• System-design interviews got a real whiteboard, and your diagram is graded as part of the interview.
• Clearer feedback reports, including a score explainer and an accurate summary card.

Older changes are reflected throughout this documentation rather than listed here.