1.1B Rejections and the Gap Line: What Mobley v. Workday Exposed

If you are over 40, have a career gap, or watched a rejection email arrive before your coffee finished brewing, the June 22, 2026 ruling in Mobley v. Workday just handed you the closest thing to a spec sheet for how AI screeners score you. Judge Rita F. Lin let the case proceed, and Workday's own filings say roughly 1.1 billion applications ran through its tools since September 2020. That is the denominator. This piece is about the numerator: the specific resume lines that plaintiffs allege get flagged as age proxies, and what to change before you hit submit again.

What Mobley v. Workday actually established on June 22, 2026

The ruling did not find Workday liable. It found that the plaintiffs' theory - that an AI screener trained on incumbent-employee data can learn to reject applicants over 40 through proxies like employment gaps, without anyone coding "reject old people" - is legally coherent enough to test. Judge Lin denied Workday's motion to dismiss the California FEHA claims and a proxy-discrimination disability claim. The collective covers applicants 40+ who applied via Workday since September 24, 2020.

Two details from the file matter for your resume:

• Workday represented that 1.1 billion applications were rejected through its software during the covered period, and told press it processes almost 1 million applications a day in 2024.
• Lead plaintiff Derek Mobley, a Black IT professional over 40 with a disability, was rejected from more than 100 positions, with emails often arriving within an hour of submission and outside business hours. That timing is the plaintiffs' central circumstantial proof that a human never touched the file.

One thing the ruling did not do: force Workday to hand over its bias-testing data. A separate order found the attorney-client privilege protected the data because Workday's counsel curated it. Translation: you will never see the model card. The proxy list has to be inferred from filings and adjacent research, and your resume defense has to be probabilistic, not surgical.

The age-proxy list plaintiffs and researchers keep naming

Court filings and adjacent hiring-AI research point at a consistent set of signals an untuned model can learn to treat as "old." None are on a Workday spec sheet. All are things you can edit tonight.

1. Employment gaps longer than 6 months, especially unexplained.
2. Graduation year more than 20 years in the past.
3. Tenure sums over 15 years across your work history.
4. Legacy technology keywords: Lotus Notes, COBOL, and other pre-2005 stacks listed without modernization context.
5. Email domain: @aol.com, @hotmail.com, and other pre-2005 providers.
6. Two-page layouts with 1998-era formatting: "References available upon request," full home address, objective statement.
7. Every-job-since-college listings that expose a 25 to 35 year timeline.

The mechanism is the one Amazon publicly walked into and abandoned: a recruiting tool trained on a decade of mostly male hires learned that maleness correlated with getting hired and started penalizing the word "women's." Nobody told it to. It read the training data. Swap "male" for "recently hired" and you have the age-proxy story: a model trained mostly on employees hired in the last five years learns that a 22-year-old graduation date and a 3-year gap correlate with "not one of us."

Illinois has already recognized this shape of harm in a different direction, specifically banning zip code as a hiring proxy because it stands in for race and class at once. "Proxy" is a legal category now, not a theoretical one.

What Refolk's index says about how few people currently own the gap

Almost nobody publicly identifies their gap, which is exactly why models trained on public-profile data have so few positive examples of "career break to hired." In Refolk's index, the visible-gap population is under 1% of a single common IC role.

That 33x ratio is the honest picture of the re-entry market. For every person publicly signaling they are on a returnship (a paid, structured re-entry program), there are 33 signaling a career break with no formal pathway attached. Models trained on this data will pattern-match "career break" to "unresolved," not "returnship graduate."

That does not mean lie. It means the way you frame the gap on the resume matters more than the fact of the gap.

Why "explain the gap in your cover letter" is bad advice in 2026

Under Workday-style screening, the cover letter often is not scored, or is scored separately after the resume clears a threshold. If the resume gets rejected in the first pass, the letter is never opened. The minute-fast rejections documented in Mobley filings are the proof: no human, and no cover-letter model, could have read the letter in that time window.

The fix is to move the gap explanation into the resume body, in fields the parser actually reads. A dated line item in your work history, titled and quantified, is scored. A paragraph in a cover letter PDF is often not.

Here is the shape that survives parsing:

• 2022 to 2024. Independent Consultant, Self-employed. Delivered 3 client engagements in fintech data pipelines. Kept certifications current: AWS Solutions Architect (renewed 2023), dbt Analytics Engineer (2024).
• 2023 to 2024. Family Caregiver Sabbatical. Full-time caregiver for immediate family. Maintained technical currency through 4 completed courses in Rust and distributed systems; contributed to 2 open-source projects (links in profile).

Both entries are dated, both name a role, both quantify something the parser can hash into features. Neither hides the gap. This is the exact rewrite work Refolk does when you paste a posting: it pulls the roles and dates out of your own history, and rewrites the gap year as a scored line item instead of a silent hole in the timeline.

Do not remove dates. Do field hygiene instead.

A 2017 Ghent University study on age-blind hiring found that concealing date of birth led to overall lower ratings, not higher. A resume with no graduation year, no work dates, and a suspicious two-column PDF layout reads to a parser as low-quality data, not as anonymous. The parser expects those fields populated. Empty fields drop your score.

Field hygiene, not deletion, is the move:

• Keep graduation year. If it is more than 20 years ago, drop the graduation year on secondary degrees and certifications, and keep it on the primary one so the parser has an anchor.
• Cap work history at 15 years of detailed entries. List older roles under "Earlier Experience" with company and title only, no dates.
• Standardize date formats (MM/YYYY). Mixed formats trip parsers and lower confidence scores.
• Use a current email domain. @gmail.com is safe. @proton.me, @icloud.com, and any custom domain are safe. Retire the @aol.com address for job applications.
• Kill "References available upon request," the full home address, and the objective statement. All three are 1998 formatting tells.

The keyword layer: modernize the stack list without lying

If your last 15 years touched Lotus Notes, COBOL, or a mainframe, do not pretend otherwise. Pair the legacy keyword with the modernization context the model has seen in "hired" examples. "COBOL on z/OS, migrated to Java Spring Boot microservices on AWS (2022)" reads very differently from "COBOL" alone. The first is a modernization engineer. The second is, to a naive model, someone who has not left 1998.

The same logic applies to titles. "Programmer" is a 1990s title. "Software Engineer" or "Backend Engineer" is the 2020s equivalent for the same work. If your title on paper was "Programmer III," you can render it as "Programmer III (Software Engineer)" without inventing anything.

This is where per-posting tailoring beats a static resume. Every job description encodes the vocabulary the screener at that company is tuned on. Refolk reads the posting, matches it against your actual history, and rewrites the phrasing to match the vocabulary the parser expects, without inventing experience you do not have. It also scores how well you actually fit, so you stop spending nights on postings where the mismatch, not the gap, is the problem.

The "agent" doctrine: who you can complain to now

If you have strong evidence you were auto-rejected in minutes for a role you were qualified for, the employer, not just the vendor, is a valid complaint target. Judge Lin's earlier rulings established that an AI tool screening on the employer's behalf can be treated as the employer's agent under FEHA and ADEA. The June 22 order kept discrimination claims moving across race, sex, age, and disability on that foundation. That exposure now reaches the 10,000+ employers using Workday's hiring stack, not just Workday itself.

Practical version:

• The opt-in window for the Mobley age-discrimination collective closed on March 7, 2026. You cannot join it now.
• You can still file an EEOC or state-agency charge on your own facts. The clearest precedent is iTutorGroup, which settled a similar EEOC case in 2022 for $365,000 after allegedly programming its AI to auto-reject women 55+ and men 60+. Over 200 applicants were involved.
• Workday was ordered to identify employers who enabled its HiredScore AI screening feature, with the list going to a third-party administrator. Ask in interviews which ATS and screening tools the employer uses. It is a legitimate question in 2026.

What to do instead of the classic "hide it" playbook

The classic advice (remove dates, drop old jobs, hide the gap) is losing to parsers that treat missing fields as low-confidence signals. The 2026 playbook is different:

1. Own the gap as a dated line item. Consulting, caregiving, sabbatical, health, study. Title it, date it, quantify one thing inside it.
2. Attach a re-entry credential where you have one. Path Forward, iRelaunch, MARS Returnship, Mom Relaunch, and Tech-Moms are the named programs recruiters recognize. A completed returnship reads to a screener as "already re-onboarded."
3. Modernize keyword pairs so legacy tech reads as depth, not staleness.
4. Match the posting's vocabulary before you submit.
5. Time your submission inside business hours in the employer's timezone. A rejection that arrives at 2 a.m. is evidence in the file. A rejection that arrives at 11 a.m. is harder to prove was automated.

The Mobley discovery does not let you see the model. It does let you see the shape of what the model is looking for. That is enough to rewrite the seven lines that are costing you interviews.

FAQ

Can I still opt into the Mobley v. Workday collective?

No. The opt-in window for the age-discrimination collective closed on March 7, 2026. What is still open is your own EEOC or state fair-employment charge if you have facts suggesting you were auto-rejected on a protected basis. The iTutorGroup settlement ($365,000 in 2022) is the cleanest precedent that AI proxy discrimination is actionable on its own facts, not just as part of a class.

Does removing my graduation year actually help?

Usually not. The 2017 Ghent University study on age-blind hiring found concealing date of birth led to lower overall ratings, and modern parsers treat empty date fields as low-quality data. The better move is to keep your primary degree's graduation year, drop it on secondary certifications, cap your detailed work history at 15 years, and list older roles under "Earlier Experience" with title and company only. Field hygiene beats field deletion.

Should I explain my career gap in the cover letter?

Not as your primary strategy. Rejections in Mobley filings often arrived within minutes, which means the cover letter was never in the loop. Move the gap explanation into a dated line item in your work history so the parser actually scores it: title the period, date it, and quantify one thing you did inside it (a certification, a consulting engagement, an open source contribution, a caregiving arc). Keep the cover letter for the human who reads round two.

How do I know if an employer uses Workday's HiredScore feature?

Ask. In 2026 it is a reasonable interview question, especially after Mobley: "What ATS and screening tools do you use for initial review?" Workday was ordered to identify employers who enabled the HiredScore AI feature to a third-party administrator, so the list exists. You can also watch the timing of rejections: minute-fast, off-hours rejection emails are the exact circumstantial pattern the Mobley plaintiffs used as evidence that no human touched the file.