Ashby's AI Resume Review: The Percentage Reviewers Can Rank By

Ashby’s page about its AI says that in application review the AI “never ‘ranks’ or gives numerical ratings to applicants”. Ashby’s knowledge base, the manual for employers who use the product, tells reviewers they can add a column showing what percentage of the criteria each applicant met and “sort by this column to rank candidates”. A separate Ashby feature, which searches an employer’s past candidates, filters out by default those whose percentage is below 51%.

We read the two pages on the same day. They can be reconciled up to a point: the AI returns a verdict on one criterion at a time, and the column is optional and ranks nothing until a person sorts by it. An applicant would still want to know the number exists.

Ashby is an applicant tracking system, the software an employer uses to collect and review applications. The employer configures it, so we can tell you what a company on Ashby is able to switch on, never what the company you applied to did switch on. Everything here comes from Ashby’s own public pages (knowledge base, website, product updates, privacy policy, API reference and trust center) as we read them on October 9, 2026. Offermic is not affiliated with Ashby. If the company you’re applying to runs a different system, we’ve given Workday, Greenhouse, Lever and iCIMS the same reading.

The employer writes the criteria

The feature in question is AI-Assisted Application Review, released on September 10, 2024. The knowledge base says it’s available on all Ashby plans and stays off until an Org Admin enables it under Opt-In Features.

For each job the employer sets the criteria the AI will check: AI-generated suggestions based on the job posting, criteria reused from other jobs, or custom ones of its own. Ashby tells the people writing them to use measurable requirements, “such as years of experience or specific previous job titles”, and adds: “Ensure the requirement can be proven by looking at a resume.” In a blog post from launch day, Ashby’s VP of People and Talent Acquisition put the applicant’s task the same way: “clearly communicate, through your resume and application, how you meet the minimum requirements in the job description.” So state how you meet the posting’s requirements where it can be verified from the page, with titles, dates and years in plain words.

A second kind of criterion checks long-text application questions, and the knowledge base is blunt about its limit: “This check only determines if the question was answered. Blank or unrelated answers are flagged as not meeting the criteria.” Answer every long-text question, and stay on its topic.

What comes back is a status for each criterion: the candidate meets it, does not meet it, or the AI was undecided. A reviewer can click any criterion to read the reasoning. The 2024 launch post called the third state “unknown” and described a fourth: “If the resume is unreadable, the criteria will be marked as skipped in the output.” Ashby’s AI page says personally identifiable information is redacted from all resumes sent to AI models, and that Ashby does not train any AI models on customer data.

The percentage, in Ashby’s own words

Four Ashby pages speak to whether this review produces a number.

Where Ashby says it	What it says
AI page, on AI-Assisted Application Review (undated, read October 9, 2026)	“The AI never ‘ranks’ or gives numerical ratings to applicants, a human must always be involved in decision-making.”
Blog post by Ashby’s VP of People and Talent Acquisition (September 10, 2024)	“No decisions are made for me. No scoring or ranking is involved.”
Knowledge base article on AI-Assisted Application Review	“To see a numeric percentage of criteria met, search for AI Job Criteria Met Percentage — sort by this column to rank candidates from best fit (highest percentage) to least fit (lowest).”
API reference, outcomeNumber field on each criterion’s evaluation	“A numerical score representing the evaluation outcome (0.0 to 1.0).”

The four agree that the AI judges one criterion at a time and that a person takes the advance or reject action. They part over the number. The percentage is the share of the employer’s criteria the AI marked as met. It’s an optional column, and the ranking happens when a reviewer sorts by it. Even so, the knowledge base tells reviewers to use it to “measure AI-based candidate fit” and speaks of “how each candidate scored against your AI criteria”.

The API reference adds a second number. It puts outcomeNumber on each criterion’s evaluation, which makes it a figure per criterion and not one per applicant, and its sample response shows 0.85 next to the outcome “Meets”. Ashby doesn’t explain how that number is calculated or whether reviewers see it.

On the decision, Ashby’s pages are consistent: the AI result does not move you by itself, and in the launch post’s words “it is up to the reviewer to advance or reject.” That step can be quick. Reviewers can filter by criteria and “Perform bulk actions on the filtered list as needed”, the AI page suggests filtering “to eliminate no-fit candidates or surface best-fit candidates”, and the launch post lists “Verify and reject candidates that meet few or none of your criteria” among the time savers. We found nothing from Ashby on how closely a reviewer has to look before a bulk action.

None of the knowledge base pages we read describes a place where a candidate can see their status, parsed profile or AI evaluation. The resume score in our resume builder is our own measure, with its rules published on that page. It is not a score from Ashby or any other applicant tracking system, and no employer sees it. The same page sets Ashby’s statement next to those of other vendors.

The bias audit has two names on it

Ashby’s AI page names FairNow as the third-party auditor it partnered with for a bias audit of its screening functionality, and the 2024 launch post named FairNow as the auditor too. The trust center carries a different name. Its public listing, “AI Criteria Evaluation - Bias Audit Report”, begins “Report prepared by WardenAI on September 21, 2026.” The launch post’s link to the audit results points to this same trust center entry, which today holds the WardenAI report dated two years later. The AI page still names FairNow, and Ashby’s pages don’t mention the change. We quote the listing only; the report itself is a PDF behind it.

On New York City’s Local Law 144, the AI page says Ashby “does not consider the currently provided tooling to qualify as an Automated Employment Decision Tool”, and that it ran the audit anyway “to support customers with differing interpretations.”

The Learn More link under the submit button

Whether you’re told about the AI review is up to the employer. Ashby’s AI page says disclosure “is the responsibility of customers to enable based on their own policies and language.” In the product it’s a setting called Automated Processing Legal Notice: “When off, candidates do not see the notice.”

When it’s on, the knowledge base says, “Candidates see a Learn More link under the Submit Application button. Selecting this link is optional. Candidates can use it to read the notice, opt out of automated processing, or check whether they have already opted out”. If you opt out, “the AI evaluation is skipped.”

An opt-out is visible to the employer. The knowledge base says “There is no proactive alert sent to customers when a candidate opts out”, though an employer can set one up, and a reviewer will “see that the job criteria evaluation was skipped, along with the reason for the skip”. The pipeline can also be filtered for candidates who opted out. Ashby doesn’t say what an employer should do with those applications. A product update dated May 28, 2026 says only that recruiters can identify them “to ensure those applications receive the right handling.”

The notice reaches only people who apply through the Ashby-hosted form. The knowledge base lists other ways in (“a direct referral, a manual add, the Chrome extension, or possibly Indeed Apply”), and a candidate who arrived by one of them never saw it. An employer can set Ashby to hide AI evaluations in that case, and reviewers then read: “The candidate has not seen the legal notice required by this job.”

Auto-Reject runs on the form, when you submit

The automatic rejection Ashby documents is a different feature. The employer sets up Auto-Reject per job posting, with conditions based on application questions, and the knowledge base says the rules “can only be configured to run at the point of the candidate’s application submission and cannot run at other stages in the interview process.”

Ashby’s example of its fuzzy “Similar to” option for typed answers is the word “local” in an answer matching a rule for “localization”. The employer also chooses how many business days to delay the rejection email. So read every screening question twice: this is the documented rejection that depends on what you typed or picked, and no person is involved. Because of the delay, the timing of a rejection email tells you nothing about whether someone read your resume.

The Auto-Reject article describes no condition that reads the resume file. If you’ve come across the claim that software rejects 75% of resumes, we traced the oldest sources we found for that number separately.

Application Limits block repeat applications. On the Legacy Plus, Plus and Enterprise plans, an employer can limit how many times a candidate applies within a scope over a number of days, or stop a rejected candidate from re-applying to the same role or to any role in a scope. Ashby identifies a prior application by the email address entered on the form, and the limits “do not apply to applications submitted via Indeed.”

What a blocked applicant sees is a message the employer can write. If the employer leaves that field blank, “the candidate sees the message from the After Candidate Submits Application field”, which another article defines as “The message that candidates see after applying for a job on your job board.” So a blocked applicant can be shown the same on-screen message as an accepted one, though Ashby doesn’t say this in so many words.

Spam protection rests on Ashby’s judgment that a submission may come from a bot. An Organization Admin picks the level, and the knowledge base warns that “A strict setting might cause Ashby to reject applications from actual candidates.”

Whether you hear back is the employer’s choice too: confirmation emails go out automatically only if the employer has set them up, and a rejection email after a human review is a toggle the reviewer turns on.

Parsing doesn’t replace your file

A reviewer on Ashby reads the file you uploaded. As the knowledge base describes the bulk review screen, it has tabs for all the application information, for the form answers alone and for the resume alone, and the candidate profile has a Resume tab. So your own layout and wording sit next to your form answers, and a parsing mistake does not replace the file.

Ashby publishes no list of the fields its parser fills. Where the employer has switched on autofill for its job board, a candidate can upload a resume and Ashby populates the application fields with its details. Those fields are in front of you before you submit, so check them: reviewers see your form answers beside the resume, and Auto-Reject reads those answers. An opt-in AI feature, Enhanced Resume Parsing, “improves parsing for resumes with more complex layouts or formatting.” Ashby doesn’t say what the standard parser does with such layouts, and it documents no way for an applicant to tell which parser an employer uses.

Resumes can be up to 50MB, and the knowledge base says they “need to be 16MB or smaller to be processed for full text searches and auto-filling application fields.” File types are vaguer: for file upload questions it names PDF, DOC and DOCX as examples of document types, and it publishes no separate list for the resume field.

For resumes a recruiter forwards to Ashby’s import email address, the knowledge base says “The file must be a PDF”, and one listed reason for failure is a file that is “an image or GIF, rather than a text-based PDF (for example, a scanned document where someone photographed a page and converted it to PDF)”. Ashby doesn’t say whether the application form’s parser has the same limits. A text-based PDF or Word file of 16MB or less stays clear of both limits. Leave off password protection too, which is one listed cause of the warning reviewers get when Ashby’s malware scan can’t finish.

After you submit, the knowledge base’s answer to whether candidates can refresh a resume or add a new one themselves is “No.” The employer has to receive the new file by email and replace it manually. Your own route is another application, since a candidate “can provide a new resume if they are applying for another job via your job board”, and that file also becomes the resume on your candidate profile.

The file travels past the first screen as well: the knowledge base says a person asked for feedback without an interview “can see the candidate’s resume, a briefing, and some basic candidate details.” To prepare for that, our resume to interview questions tool lists the questions your own resume is likely to raise, and our guide to resume-based interview questions covers what interviewers ask from a resume.

A profile that outlasts the application

One candidate profile can hold applications to different roles, and reviewers can see the others. An application joins a profile by email: “If the email used to apply matches an existing candidate profile, Ashby adds the application to that existing candidate — no duplicate is created.” Apply with different information and “an additional candidate profile is created but may be flagged as a duplicate.” The duplicate check looks at email addresses, name and social links, close name matches included.

Applying twice to the same role with the same information, while the first application is still active, leaves one job consideration on your profile, and the employer sees both forms. Because the application limits and the duplicate check both turn on your email address, use one address per employer. A second one can produce a flagged duplicate profile, not a fresh start.

After a rejection, your profile stays in the employer’s Ashby account until the employer removes it, and this is where the percentage comes back. AI Talent Rediscovery, an opt-in feature on the Legacy Plus, Plus and Enterprise plans, lets an employer search its past candidates for a new role. Ashby lists five negative signals “causing candidates to be filtered out by default”: “Recently Rejected (last 6 months)”, “Poor Fit”, “Poor Interview Performance”, “Candidate Ghosted” and an AI Job Criteria Met Percentage below 51%. That last one is a numeric cut-off on the AI result. It applies to this search of past candidates, not to new applications.

How long the profile stays is the employer’s setting. The knowledge base mentions default retention rules without giving a default period, and says the rules only mark records: “Ashby flags candidates who violate your data compliance rules, but it does not anonymize or delete them automatically.” An employer that acts can anonymize the record, which permanently removes PII from it, or delete it. Ashby says a deleted record is permanently erased from its application database after 10 days, and from its backend when a 30-day backup retention window expires.

A request to see, correct or delete your data goes to the employer. Ashby’s privacy policy says its customers are the data controllers and recommends that you “reach out directly to the company you applied to”.

Before you press Submit Application

A lot is fixed at the moment you submit. Auto-Reject rules run then and at no other stage. The resume you attach stays with that application, and Ashby’s answer on replacing it yourself is “No.” Where an employer shows the AI notice, the link that lets you opt out sits under that same button.

So the time to act is before the click. Read the screening questions twice and check what autofill put in the fields. If there’s a Learn More link, open it and decide about the opt-out, knowing the employer can see your choice. Afterward, the documented ways to change your resume or your data run through the employer.