In May 2025, we published a list of safer words. Say “welcoming culture” instead of the acronym. Say “resilient communities” instead of “climate justice.” It was the advice everyone was giving, and we gave it too.
A year and a half of enforcement data later, we’d write it differently. So here’s the update, because the swap isn’t doing what most organizations think it’s doing, and a lot of people made real decisions on the strength of that assumption.
What we got right
Your values didn’t change. The language you use to talk about them did have to.
That part holds. The vocabulary that read as “we’re one of the good ones” in 2021 now carries exposure it didn’t carry then, and pretending otherwise doesn’t serve anybody. If you rewrote your site in the last two years, you weren’t being cowardly. You were reading the room.
What we got wrong
We treated it as a vocabulary problem. It’s two problems, and the swap only touches one of them.
Problem one: your words are being machine-read. This is real, and it’s newer than most people realize. Court records unsealed this summer showed federal agencies running keyword searches to flag grants for review and cancellation. Not reading each application. Searching it. The National Science Foundation (NSF) list included “advocates,” “biases,” “historically,” “inclusivity,” “clean energy.” Grants were pulled en masse on the strength of a term appearing. Some were reinstated by court order months later, after the organizations had already absorbed months without the funding.
Against that kind of screening, changing a word does change the outcome. The software isn’t weighing whether your program is fair. It’s looking for a specific term and flagging whatever page contains it. Take the term out, and the page stops getting flagged. That’s the one place the swap genuinely works, and it’s worth saying, because it’s the reason so many organizations believed the swap worked everywhere.
Problem two: enforcement looks at conduct, not labels. In this scenario, the swap does nothing at all.
The U.S. Equal Employment Opportunity Commission (EEOC) said so in as many words in February 2026 guidance, which extends its scope to “functionally similar programs, policies, and practices that now may bear other labels, such as Inclusion & Diversity; Belonging; People & Culture; or Opportunity & Inclusion.”
If you read that list again, you will notice that it is—almost word for word—what every word swap of the last two years landed on.
The Justice Department was blunter. In April 2026, IBM paid $17 million to settle claims that it discriminated in hiring and promotion. What got IBM in trouble wasn’t the name of its program. It was tying executive bonuses to hitting demographic numbers and limiting who could get into certain training programs. Announcing the settlement, the acting attorney general said government contractors “cannot evade the law by repackaging it.”
Two more pieces of the same machinery: a March 2026 executive order requires federal contractors to certify they aren’t running discriminatory programs, treats that certification as material to payment under the False Claims Act, and flows the obligation down to every subcontractor. And since early 2025, ten state attorneys general have had an open inquiry into major financial institutions’ diversity and ESG programs.
None of those enforcement actions look at your website. They look at how you hire, promote, pay, and train. So renaming a program doesn’t change the outcome. The program is still the program.
That’s the correction. The swap protects you from one narrow thing: an automated search for a word. It does nothing about the rest. We wrote it in 2025 as though it covered both, but the last year and a half tells us that it doesn’t.
Who this actually applies to
Here’s the thing that should stop most readers of this post in their tracks.
Most of that machinery (the False Claims Act exposure, the certifications, the proxy fights, the attorney general inquiries) is aimed at federal contractors and public companies. If you’re neither, none of it reaches you. Worth checking whether you subcontract to anyone who is, though, because those obligations flow down.
What does still apply, to any employer with fifteen or more people, is Title VII itself: how you actually hire, promote, pay, and train. Which is the same point as before. The conduct is what’s regulated, not the vocabulary.
So for a privately held purpose-driven business under $250 million with no federal contracts: no False Claims Act exposure, no proxy fight, no attorney general inquiry, no grant portfolio to lose.
And they’re scrubbing anyway.
The numbers on the corporate side are remarkable. In 2024, 90% of S&P 500 companies used the language in their disclosures. By early 2026, 8% did. Diversity-related disclosure went from 97% of the index to 55%. That’s a stampede, and it was driven by legal exposure that publicly traded federal contractors genuinely have.
What we watch happen next is smaller organizations copying that posture without the exposure that produced it. They see the big companies go quiet, read it as the new normal, and delete the two sentences that were doing all the differentiating work on the site. Anticipatory self-censorship, in response to a risk that isn’t theirs.
Nonprofit brand strategy is a different calculation
Nonprofits are in a different position: if you hold federal funding, the scan is real, the stakes are your budget, and the calculus is not the same one a private company is running. The difference isn’t nerve. It’s exposure. Know which one you have before you decide.
What actually works: brand messaging strategy, not vocabulary
Not a substitution list. Four things:
- Describe conduct, not categories. “We hire from the neighborhoods we serve” survives both problems. It’s not a flagged term and it’s not a renamed program. It’s a true sentence about what you do. Abstractions are what get scanned and what get scrutinized; specifics are neither.
- Put the weight on narrative, not vocabulary. A term like “equity” was doing the work of an entire program. When the word became a liability, the story went with it, because there wasn’t a story underneath. Just a label. Organizations built on narrative haven’t had that exposure. That’s the execution layer we’ve written about separately.
- Say the plain version instead of saying nothing. Silence is the most common response we see and the most expensive one. It reads as absence, and absence reads as “not sure they actually stand for anything.”
- Decide on purpose. Name which of the two risks you actually carry (the scan, or the conduct), decide what you’re willing to trade for it, and write it down. Most organizations have made this trade in a hundred quiet edits without anyone ever deciding. The edits aren’t the problem. The absence of a decision is.
What going quiet costs
An organization that goes quiet doesn’t become neutral. It becomes indistinguishable from every other nonprofit and purpose-driven business that also went quiet, which is most of them, this year.
Plenty of organizations share your cause. What sets you apart is the specific thing you do inside it: who you serve, how you work, what actually changes because you exist. That’s real whether or not anyone can see it.
Language is how people find out. Strip enough of it away and the difference doesn’t disappear. It just stops being visible. And from the outside, invisible and identical look the same. You’re one more beige tab in a browser full of beige tabs.
The organizations doing this well right now aren’t the loudest ones. They’re the clearest ones: the ones who found the plain, true sentence underneath the buzzword and said that instead.
Start with a messaging audit
Find out what your language is actually doing.
Messaging Audit: $5,000 | two weeks.
We read the public-facing content you have: site, collateral, grant boilerplate, LinkedIn, the about page nobody’s opened since 2023. You get back what your language is currently saying, which of the two risks actually applies to an organization your size, and a specific list of what to change and what to leave alone.
If what surfaces is bigger than language, and sometimes it is, we’ll tell you that instead of selling you the audit twice. That’s strategy work built for organizations like yours, and it’s a different conversation.
Sources
- U.S. Equal Employment Opportunity Commission, “Reminder of Title VII Obligations Related to DEI Initiatives”, February 26, 2026. The “other labels” language.
- U.S. Department of Justice, “IBM Pays $17 Million to Resolve Allegations of Discrimination Through Illegal DEI Practices”, April 10, 2026.
- Higher Ed Dive, “Ctrl+F: Federal agencies used keyword searches to nix grants en masse, court records show”, July 23, 2026.
- Harvard Law School Forum on Corporate Governance, “Evolving Legal and Regulatory Dynamics for DEI Challenges and its Impact on Corporate Disclosures”, June 8, 2026. The S&P 500 disclosure figures.
- Jackson Lewis, “New EO Targets Federal Contractor DEI Practices”, on the March 26, 2026 executive order.
- Jackson Lewis, “Ten State Attorneys General Launch Inquiry into Major Financial Institutions’ DEI & ESG Programs”, January 29, 2025.
- Similarweb, “Zero-Click Marketing: What the 2026 Data Means”. 68% of Google searches ended without a click as of mid-2026.