What an Alex (Apriora) AI Interview Records and Scores

If a recruiter invites you to an interview with "Alex," you are being screened by a product from Apriora Inc., which now does business as Alex. Its own employer-facing documentation states what the session produces: "transcripts, role-fit scores, & fraud flags directly to your ATS," per the Alex interview reports page. Two of those three outputs are never mentioned on the page Alex shows candidates. This article explains what is generated around an Alex interview, which parts start before you speak, and what you can reasonably do about each.

All product descriptions below are quoted from Alex's own marketing and documentation pages as they read on 12 September 2026. They are vendor claims about vendor capabilities, not independent test results.

Who Alex is, and why the name changed

Alex is an AI recruiting product built by Apriora Inc., a Y Combinator Winter 2024 company founded by Aaron Wang and John Rytel. The Y Combinator company profile describes it as a system that "interviews, screens, and schedules every candidate – over voice, video, and text," and claims "past 1M+ interviews" and integrations with "33+ ATS platforms."

The product was marketed as Apriora until the company adopted the Alex name; apriora.ai now redirects to alex.com, and the alex.com footer still reads "Copyright © 2026 Apriora Inc. dba Alex." If you received an invitation referencing either name, or a calendar link from a domain you did not recognise, they are the same system. This matters mainly because searching for the wrong name returns almost nothing useful.

The three artifacts an Alex screening produces

The interview reports page is the clearest statement of output. It says recruiters receive "searchable transcripts, role-fit scores, and fraud flags, so you can make hiring decisions based on verified skills," and that these land in the applicant tracking system rather than in a standalone tool. It also notes recruiters can "Chat directly with Alex with follow-up questions" about a candidate.

Three practical consequences follow from that.

Your words become durable, searchable text attached to your record. A transcript in an ATS is not a recruiter's memory of a call. It persists, it is greppable, and it can be read months later by people who were not in the conversation and who lack the context of how the question was phrased.

A score accompanies it. The Alex homepage describes evaluation using "standardized questions and objective rubrics." "Objective rubric" is a vendor characterisation, not a verified property; what it means concretely is that a number or band travels with your transcript and may be the first thing a recruiter sees.

A fraud signal travels with both. This is the output candidates are least likely to expect, and it is worth treating separately.

Verify: a fraud band assigned before the interview

Alex sells a pre-interview product called Verify. Its product page states that "Alex reads the email, the phone, the connection and the professional footprint together, and sets a fraud risk band on each candidate," and that "Each candidate carries a fraud risk band, low, medium or high, with the signals behind it open for you to read."

The signals the page names include email disposability and age, whether the phone number is registered as VOIP, whether the connection arrives through a VPN or proxy, the LinkedIn profile, and whether professional history depth matches the résumé timeline.

The page is explicit that none of this is visible to you: "No extra steps for the candidate, nothing new for your team," and "The candidate never sees a step that was not part of applying."

Read plainly, that means ordinary privacy habits are scored inputs. A Google Voice or other VOIP number on an application, a privacy-forward email provider, a thin LinkedIn presence, or a VPN you leave running by default are each things the documentation says contribute to a band an employer reads before anyone speaks to you. None of them indicate fraud, and Alex does not claim they individually do; the band is a composite. But if you are deciding which phone number to put on an application, the vendor has told you which one reads cleaner. If you want the broader picture of what connection and location data exposes during a remote interview, see how remote interviews reveal your location.

Identity: a government ID matched against a live selfie

Separately, Alex offers Identity, which its product page says "checks a government-issued ID against a live selfie while the interview is happening." The page says the check looks for altered or rebuilt documents, photographs of a screen, expired documents, whether the ID photo matches the live capture, and whether the name is consistent across the document and the application.

It also states the capture must be live: "A live capture is matched against the photo on the document. A saved photo, an uploaded file, or an image held up to the camera cannot stand in for the person." A verified result appears as "an identity-verified badge on the report," and Alex says "The decision stays with your team."

Two things follow. First, if an employer has this module enabled, expect to be asked for photo ID mid-interview, which is a poor moment to discover your licence is in another room. Second, name mismatches are a named failure condition, so an application filed under a preferred or shortened name that does not match your document is a foreseeable source of friction worth raising with the recruiter in advance rather than mid-session. Our candidate guide to interview identity verification covers how these checks typically work across vendors and what to ask before you submit a document.

What the candidate-facing page tells you instead

Alex's candidate page is built around convenience and sentiment. It emphasises interviewing "24/7," instant scheduling, and "real-time feedback," with testimonials describing the experience as "surprisingly human." As of 12 September 2026 it does not state that the session is transcribed, that a role-fit score is produced, that a fraud risk band exists, or who reads any of it.

That gap is the reason this article exists. Nothing on the employer pages is hidden — they are public — but the asymmetry means candidates typically learn what an Alex interview produces only if they go looking at pages written for the people buying the product.

Whether you are entitled to more than that depends on where you are interviewing. The EU AI Act's transparency duty and the US state and city notice rules differ substantially in what they require an employer to tell you, and we cover the current state of those obligations in what AI hiring rules entitle you to know in 2026.

What Alex does not document publicly

Being candid about the limits of this reading:

• Retention. Neither the product pages nor the company's GDPR compliance announcement from December 2024 states how long transcripts, recordings, or identity captures are kept. That post describes rights "to access, correct, or delete" personal data but gives no retention period.
• Recording versus transcription. The pages consistently promise transcripts. They do not clearly state whether raw audio or video is retained alongside the transcript, or for how long.
• Score mechanics. "Role-fit score" and "objective rubrics" are not defined publicly in terms of inputs, weighting, or thresholds.
• Controller. In most deployments the employer, not Alex, decides retention and access. The employer is the party to ask.

If those answers matter to you, the employer's recruiter is the correct recipient of the question, and asking is not an unusual request.

How to prepare for one

• Treat it as an on-the-record screening, because it is one. Answer the question asked, in full sentences, as though a hiring manager will read the sentence out of context later — because a searchable transcript is exactly that.
• Use the real phone number you actually answer, and apply from a normal connection rather than a VPN endpoint in another country, unless you have a reason to prefer otherwise and are comfortable explaining it.
• Have government photo ID within reach before you join.
• Check the name on your application against the name on your ID, and flag any mismatch to the recruiter beforehand.
• If a human round matters to you, ask for one explicitly rather than assuming the AI screen is a formality.

Where a candidate-side assistant does and does not fit

Control is a desktop AI interview assistant for Windows and macOS, built for screen-share and coding workflows where the useful help is reading what is on screen. An Alex screening is mostly a spoken conversation about your background, often with an identity check attached — a format where the honest answer is that preparation beats live assistance, and where any tool that reads from a screen has little to work with.

The more useful preparation for a conversational AI screen is knowing what the system produces and rehearsing answers that survive being read as text. If you want to see how other AI interviewers in this category handle recording and scoring, what a Mercor AI interview records covers a comparable product with more published detail about who reviews the output.

The short version

An Alex interview produces a searchable transcript, a role-fit score, and a fraud risk band, all delivered into the employer's ATS. The fraud band is set before you speak, from signals including your email, phone type, connection, and professional footprint, and Alex states candidates never see that step. If the Identity module is enabled, you will be asked to match a government ID against a live selfie during the session. Retention periods and scoring mechanics are not published; the employer is the party that can answer both.