# The Outreach Pay-Disclosure Standard: Cleared, Fix, or Hold

*Grade any outreach template or screen script Cleared, Fix, or Hold against the pay rules where its candidates live, so a second reviewer reaches the same grade.*

- Canonical URL: https://www.refolk.ai/guides/outreach-pay-disclosure-standard
- Pillar: Process, data, and compliance
- Format: Standard
- Published: 2026-10-29
- Last reviewed: 2026-10-29
- Reading time: 20 min
- Keywords: do recruiter messages need a salary range, pay transparency rules for sourcing outreach, can recruiters ask current salary on screening call, salary history ban recruiter script, pay range disclosure remote candidates by state, EU pay transparency directive recruiter outreach

## Key takeaways

- Treat any outreach template sent to more than one person as a job advertisement, because New York State's proposed regulations count a message to a pool of more than one prospective applicant.
- A remote sequence is graded against every location it reaches: New York covers roles reporting to a New York supervisor, office, or work site, and Colorado covers remote postings even when they exclude Colorado applicants.
- Each pool message needs a dated, good-faith range for one location and level, and a narrow spread proves nothing on its own: NYC challenged listings at $150,000 to $225,000 as not in good faith.
- A screen script may ask about pay expectations but never current or previous pay, and a soft version like asking whether the candidate would like to share current pay still counts as prompting.
- In Refolk's index, 6,317 US agency or executive recruiters work in staffing, compared with 5,726 sourcer-titled people, so the pay-history bans that reach agents in California, Maryland, and Connecticut cover much of the actual asking.
- Penalties for missing ranges run up to $250,000 in NYC after a 30-day cure window and $500 to $10,000 per violation in Colorado, which reported over 2,800 complaints in one enforcement update.

This is the sign-off standard for whoever approves outreach templates, sequence steps, and recruiter screen scripts before a search goes live. It turns pay-range disclosure laws and pay-history bans into fixed criteria. Any script, together with the candidate locations it will reach, gets one grade: Cleared, Fix, or Hold. A second reviewer working from the same records should reach the same grade. This is a policy standard, not legal advice. Where the law is unsettled I say so and set a conservative default.

## The three grades and what each one requires

Every item gets exactly one grade:

- **Cleared:** it ships as written to every mapped location.
- **Fix:** it has a defect the reviewer can correct from records already on file.
- **Hold:** it is blocked by a missing input or an open question that only the compensation team or counsel can close.

An item is one outreach template, one sequence step, or one screen script. Each item is graded against the full list of jurisdictions its recipients sit in.

| Grade | Meaning | Typical cause | Next owner |
|---|---|---|---|
| Cleared | Ships as written | All elements present, reviewers agree | Sourcer |
| Fix | Correctable now | Missing benefits line, "current pay" wording | Reviewer |
| Hold | Blocked | No dated range, agency script not collected | Comp team or counsel |

Three tie-breakers keep two reviewers on the same grade:

- **The worst jurisdiction sets the grade.** An item that is Cleared for Washington and Fix for Colorado is graded Fix.
- **A missing input is always Hold, never Fix.** A reviewer cannot make up a range they were never given, or approve a script they have not seen.
- **Disagreement means Hold.** If the reviewers still disagree after talking it through, the item is Hold until the disagreement is written down and resolved.

> **Rule:** Grade against the strictest location
>
> An item is graded against every jurisdiction its recipients could trigger, not just where the hiring manager sits. A sequence that passes in nine of ten locations is not Cleared.

#### The evidence under a Cleared grade

1. **Second-review grade** - Two reviewers grade separately, reach the same result, and record it
2. **Screen script** - Expectations questions only, including agency and AI-drafted versions
3. **Message disclosures** - The range plus every extra element the mapped jurisdictions add
4. **Dated range record** - One good-faith range per location and level, with date and rationale
5. **Jurisdiction map** - Every candidate location tagged against the reports-to and could-work-there tests

*A Cleared grade is only as strong as its weakest layer, and a missing range record makes every message built on it a Hold.*

## When does a recruiter message count as a job advertisement?

Treat any outreach template that goes to more than one person as a job advertisement. New York State's proposed regulations count an ad sent to a "pool of applicants," which means more than one potential or prospective applicant. They name "an e-mail sent to a pool of applicants" and "an e-communication sent using an electronic mailing list" as examples.

That is the only clear test in the public record, and it settles most sourcing work:

- **Sequences are pools.** A sequence step loaded with 40 names is one advertisement sent 40 times, however many merge fields it uses.
- **Bulk InMail and mail-merge campaigns are pools.** The test is the same: count the recipients per template.
- **A "one-to-one" template that gets reused is a pool.** Once the same body text goes to a second person, the template is an ad.

**What the test proves:** that a regulator applying New York's wording would see an advertisement. **How it misleads:** a message can feel personal because it has a first-name token and a line about the candidate's repo. Personalisation changes the tone. It does not change how many people received the template.

Recruiters are clearly within scope. New York's complaint form lets the person complaining name the poster as "Employer, Job recruiter, Job posting website." The proposed rules excuse employers for third-party ads posted without their knowledge, such as scraped ads. A message your own sourcer sent is the opposite case.

Colorado adds a rule about format. An e-mail meets the requirement "provided that the information is hyperlinked or accessible via a URL within the electronic posting." So in Colorado, a link to a live posting that shows the range can carry the disclosure. Whether New York's similar allowance made it into the final rule is unverified, and the final New York rule text itself is unverified. Check the New York DOL Pay Transparency FAQ before relying on either.

> **Note:** What is not established
>
> No regulator has published guidance on sequenced InMail or on a truly one-to-one message, and I found no enforcement case about outreach messages at all. My default is that a one-to-one message naming a specific open role still carries the range. It costs one line, and the alternative is betting on an exception nobody has written down.

> Personalisation changes the tone of a message, not the number of people who received the template.

## Which candidate location sets the rule for a remote search?

Every location the search reaches can set a rule, and the strictest one decides the grade. New York asks where the role reports. Colorado asks where the work could be done. A single remote sequence can trigger both at once.

The two tests:

- **New York (reports-to test).** A posting is covered "if the employee or applicant is reporting to a supervisor, office, or work site in the state of New York." A remote engineer who reports to a manager in New York is covered wherever they live.
- **Colorado (could-work-there test).** A remote posting is covered "even if it states that the employer will not accept Colorado applicants." Writing "no Colorado" into the message does not remove the obligation.

#### Which remote-role test fires

Horizontal axis runs from Reports outside New York to Reports to a New York supervisor or site. Vertical axis runs from Work cannot be done from Colorado to Work could be done from Colorado.

| Quadrant | What it means |
| --- | --- |
| Colorado only | Colorado-specific range, bonus, commission, benefits, and application deadline |
| Both tests fire | New York base range and job description, plus every Colorado element |
| Neither test fires | Still check each candidate's own state against the jurisdiction list |
| New York only | Base-pay range for that location plus a job description |

*A remote role that reports into New York and could be done from Colorado must meet both states' disclosure rules in the same message.*

Headcount thresholds decide whether the employer is covered at all:

- **New York State:** 4 or more employees
- **New York City:** 4 or more, and employment agencies of any size
- **New Jersey:** 10 or more
- **Washington and Illinois:** 15 or more
- **Massachusetts:** 25 or more
- **Colorado and Maryland:** all employers

The range-in-posting jurisdictions in this standard's reference set are California, Colorado, Connecticut, Hawaii, Illinois, Maine, Maryland, Massachusetts, Minnesota, New Jersey, New York, Vermont, Virginia, Washington State, and DC. City rules apply in New York City, Jersey City, Ithaca, Cincinnati, Toledo, Cleveland, and Columbus. Delaware has passed a law that takes effect later. The list keeps growing, so re-check it before you adopt this standard and then on the schedule in the last section.

Mapping is mostly a data problem. You cannot say which tests fire until every person in the sequence has a location. If the slate was built in [Refolk](/) from a query that names a place, the location was part of the search from the start. The jurisdiction list then comes straight from the query instead of from a cleanup pass on a spreadsheet.

## What range and disclosures must a pool message carry?

Each pool message carries a good-faith range set for one specific location and level, plus every extra element the mapped jurisdictions add. The range is base pay only for New York State and NYC. NYC defines good faith as "the salary range the employer honestly believes at the time they are listing the job advertisement that they are willing to pay the successful applicant(s)."

| Jurisdiction | Headcount | Max penalty | Range content |
|---|---|---|---|
| NY State | 4+ | $3,000 | Base pay plus job description |
| NYC | 4+, agencies any size | $250,000 | Base pay only |
| Colorado | All | $10,000 per violation | Range, bonus, benefits, deadline |
| Washington | 15+ | $5,000 | Range plus benefits |
| New Jersey | 10+ | Unverified | Range plus benefits |

New York State's penalties step up from $1,000 to $2,000 to $3,000. Colorado's run from $500 to $10,000 per violation, and Washington's from $100 to $5,000.

Range rules a reviewer can check without a lawyer:

- **No open-ended ranges.** "Up to $X" or "$X+" fails in NYC and Colorado.
- **One range per location.** New York's proposed rules refer to "a single opportunity and a single geographic location." Colorado expects the Colorado range, not a national blended one.
- **Base pay stands apart.** NYS and NYC want base pay only. Colorado and Washington want bonus and benefits described separately.
- **Extra elements.** Washington requires "a general description of all of the benefits and other compensation." Colorado adds bonus, commission, benefits, and an application deadline. This standard also adds benefits wording for New Jersey, Illinois, and Maryland, and a job description for New York State.

The spread between minimum and maximum is where good faith gets tested, and the ratio can mislead in either direction. NYC questioned maximums 3 to 4 times the minimum, so compute max divided by min for every range. But NYC also brought a complaint against a firm over five listings at $150,000 to $225,000, calling them not in good faith, and that is only a 1.5x spread. A narrow range proves nothing on its own. What supports good faith is the dated record of why the employer believed it would pay inside that range. One secondary source says New Jersey caps the spread at 60% of the minimum. That is unverified, so do not grade against it until you have checked the statute.

**Pay disclosure block for a pool message**

```
Pay for this role ({location}, {level}): ${min} to ${max} base salary per year.
Bonus or commission: {plain description, or "none"}.
Benefits: {general description of all benefits and other compensation}.
Apply by: {date}
Full role description and pay details: {link to live posting showing the same range}

[Internal only, strip before sending]
Range record: {range ID}, set {date} by {comp owner}
Jurisdictions covered by this block: {list}
```

*Fill one block per location and level from the dated range record, and delete any line whose jurisdiction is not on the map.*

## Which compensation questions may a screen script ask?

A screen script may ask what the candidate expects to earn. It must never ask what they earn now or earned before, however gently the question is phrased. Around 20 states plus DC ban pay-history questions for private employers, and a soft version is still a question: "would you like to tell us your current pay?" is still prompting.

The bans vary in ways that change the script:

| Variant | Jurisdictions | Script consequence |
|---|---|---|
| Ban reaches agents or third parties | California, Maryland, Connecticut | Agency and vendor scripts graded too |
| Expectations expressly allowed | Illinois, Colorado, Kansas City | Keep the expectations question |
| Confirmation allowed after offer | New York, Delaware, Maine | Post-offer only, never in a screen |
| Volunteered pay may support higher offer | Maryland, Minnesota | Never prompt; counsel sets handling |

**What a clean question proves:** the script asks for a number about the future. **How it misleads:** "What would it take for you to move?" followed by "and where are you today?" The second half is a pay-history question. Search every script for "current", "previous", "today", "now", and "last". Each hit is a Fix unless the reviewer can show it has nothing to do with pay.

When a candidate volunteers past pay without being asked, the script stops them politely and moves on. Maryland and Minnesota allow voluntarily disclosed pay to support a higher offer, but that is a decision for counsel, not something the sourcer writes into the ATS on their own initiative.

**Screen-script compensation section**

```
"Before going further: the pay range for this role in {location} is ${min} to ${max} base, plus {bonus and benefits summary}."
"What are you looking for in base salary for your next role?"
"Does that range work for you?"

If the candidate starts to share current or past pay:
"Thanks, I don't need that. What matters is what you're looking for next."

Never ask: current salary, previous salary, last bonus, current equity, "where are you today", "would you like to share your current pay".
```

*Paste this over the existing compensation section and do not add follow-up questions.*

## Agencies, AI drafting tools, and candidates in the EU

Anything asked on the employer's behalf is graded as if the employer asked it. The bans in California, Maryland, and Connecticut expressly reach agents and third parties, and NYC covers employment agencies of any size. NYC's Commission has brought complaints against "employers and employment agencies," so agencies are covered by the disclosure rules as well as the history bans.

| Group | Count |
|---|---|
| US agency or executive recruiters, staffing industry | 6,317 |
| US sourcer-titled people, any industry | 5,726 |
| Agency to in-house ratio | 1.10 (derived) |

**6,317 - US agency or executive recruiters in the staffing industry, in Refolk's index**

That is more than the 5,726 US sourcer-titled people, so a large share of the actual asking happens outside your own team.

The practical step is to collect the scripts. I don't mean the agency's policy statement. I mean the words their recruiters actually say and the templates their sequences actually send, for this search. AI drafting tools work the same way: grade what the tool produced for this search, not the prompt someone wrote for it. A drafting tool can bring back a "current compensation" line in any new draft, so each draft is a separate item. If an agency will not share its script, the item is Hold.

I ran this search: `Recruiting operations leads at US staffing firms with offices in New York City and Colorado` - [see the full result list](https://www.refolk.ai/s/732g5s5vr2).

*Returns named recruiting-ops people at staffing firms that work in both strict markets, which gives you the list of people to ask for actual agency scripts.*

### Candidates in the EU

The EU Pay Transparency Directive (2023/970) sets a minimum, and the national law that implements it (transposes it) sets the actual rule. Article 5(1) requires the starting pay or range to be given "in a published vacancy notice, prior to the job interview, or otherwise." Article 5(2) bans pay-history questions.

When the research behind this standard was done, five member states had fully transposed the Directive: Slovakia, Lithuania, Italy, Malta, and Greece. Italy, Lithuania, and Slovakia require the range in the advert itself, and Slovakia also requires it before pay negotiations begin. Lithuania required salary in job ads before the Directive existed. Seven member states had published no draft, Germany among them. How the national laws treat questions asked through intermediaries is not publicly established.

The default for this standard:

- **Candidate in a state that has transposed:** the range goes in the outreach message.
- **Candidate in a state that has not:** the range goes in the message anyway if you can, and before the interview at the latest, which is the Article 5(1) minimum.
- **Everywhere in the EU:** no pay-history questions in any script.

| Market | Sourcer-titled people | Top title | Top-employer concentration |
|---|---|---|---|
| United States | 5,726 | Sourcing Specialist | No employer above 1 in a sample of 25 |
| Germany | 178 | Sourcing Specialist | 32% of sample at two employers |
| US to Germany | 32.2x (derived) | | |

In Refolk's index, the German sourcer population is small and concentrated. In the sample, Volkswagen Group Services and Siemens together account for 32%. My inference, and it is only an inference, is that German practice will be set employer by employer before national law arrives. A German search should follow the employer's own policy, plus the Article 5(2) history ban as a minimum.

## The pre-launch grading procedure

Run these seven steps for every search before the first message goes out. Steps 1 to 6 happen before launch. Step 7 runs for as long as the search is live.

#### Grading a search's scripts Cleared, Fix, or Hold

1. **Map where candidates are** - Tag every target location against the New York reports-to test, the Colorado could-work-there test, and headcount thresholds. Done means a written list of triggered jurisdictions for the search, in about an hour.
2. **Set and log the range** - The compensation team sets one good-faith range per location and level and records the date and rationale. Done means a dated range record for every mapped location, usually in 1 to 2 days.
3. **Classify each message** - Label every template and sequence step as pool or one-to-one, counting recipients per template rather than per send. Done means every item carries a label, in about 30 minutes.
4. **Insert the disclosures** - Add the location range, bonus and benefits wording for Colorado, Washington, New Jersey, Illinois, and Maryland, a job description for New York State, and a deadline for Colorado. Done means no item is missing an element that applies to it.
5. **Scrub the screen script** - Remove every current or previous pay question and keep only expectations questions, including in agency and AI-drafted versions. Done means each script has been reviewed line by line.
6. **Grade with a second reviewer** - Two reviewers grade each item Cleared, Fix, or Hold independently. Done means both reach the same grade, or the item is held with the disagreement recorded.
7. **Monitor and cure** - After launch, check live messages and linked postings against the range record, and act inside Washington's 5-business-day and NYC's 30-day cure windows. This step continues while the search is live.

Sources disagree about order. One multi-state guide puts setting the range before writing the ad. Washington-focused counsel put the weight on checking postings after they go live. This standard does both: step 2 must finish before step 4 begins, and step 7 covers checks after launch. A message written before its range record exists is Hold by definition.

## How this standard goes wrong

Most failed grades come from a handful of predictable mistakes, and each one has a mechanical check. Run these checks on every item graded Cleared, because these are the mistakes that produce a Cleared grade that should not have been given.

- **A "personalised" sequence treated as one-to-one.** Merge fields make a sequence feel individual, but sending to more than one person meets New York's pool test. *Check:* count recipients per template, not per send.
- **Excluding Colorado from a remote role.** Colorado covers remote postings even when they exclude Colorado applicants, so "no Colorado" is not a fix. *Check:* does the message say "remote" anywhere? If it does, the Colorado elements apply.
- **The range only in a link.** Colorado accepts links. Whether New York's attachment allowance survived into the final rule is unverified. *Check:* open the link at grading time and again at launch. If the link is dead, the message has no range.
- **Ranges that are too wide or open-ended.** "Up to $X" fails outright, and NYC questioned maximums 3 to 4 times the minimum. *Check:* divide max by min, and confirm a dated rationale exists however narrow the spread.
- **The polite history question.** "Would you like to share your current pay?" is still prompting. *Check:* search every script for "current" and "previous", then read each hit in context.
- **Agency or AI script drift.** The California and Maryland bans reach agents, and a script approved last month may not be the one in use today. *Check:* collect the actual scripts agencies and AI tools use for this search, not a sample from another one.
- **A national blended range.** Colorado expects the Colorado-specific range, and New York points to a single geographic location. *Check:* one range per location on the range record, and the message uses the right one.

> **Watch out:** Sequence variants multiply exposure in NYC
>
> Colorado counts multiple postings for the same job as one violation, but NYC exposure applies to each uncured violation. Five split-test variants of a defective sequence step can mean five problems, not one.

The standard also has limits of its own, and a reviewer should know them:

- **It cannot settle one-to-one messages.** No regulator has ruled on them. The default to include the range is a policy choice, not a legal finding.
- **It does not test good faith on its own.** The ratio check catches obvious cases. The JK Partners complaint shows a 1.5x range can still be challenged, so the range record is what stands behind a Cleared grade.
- **It is weakest on intermediaries in the EU.** How the national laws treat agency questions is not established, so EU agency scripts default to the strictest reading.

## Keeping the standard current

Re-check the jurisdiction list, cure windows, and EU transposition status on a fixed schedule, because each changes by statute or ruling rather than by announcement to you. A quarterly review, plus a re-check whenever a search opens in a new state or country, keeps the standard honest.

What to watch, and where to look:

- **New jurisdictions and effective dates.** Delaware has passed a law with a later effective date, Connecticut is among the newer additions, and Virginia's pay-history ban is one of the newest. Re-check the state list against the NY DOL Pay Transparency FAQ, the NYC CCHR Salary Transparency Factsheet, and Colorado's CDLE INFO #9A for changes to the two strictest rules.
- **Washington's cure window has a sunset.** Washington's amended law added a 5-business-day cure, and that provision ends on a fixed statutory date. Separately, the Washington Supreme Court's Branson decision held that a plaintiff need not be a bona fide applicant. Once the cure window ends, claims from test applicants become harder to fix after the fact. Check RCW 49.58.110 for the cure provision's current status.
- **Colorado enforcement volume.** One published Colorado enforcement update reported over 2,800 complaints and more than $800,000 in fines. That volume is why Colorado's test usually decides the grade for remote searches.
- **EU transposition.** The list of transposed states will grow. Re-check against a public tracker such as Lockton's EU transposition tracker or the L&E Global 27-state report before any EU search, and move a country into the "range in the message" group once its law requires the range in the advert.
- **The final New York rule text.** The pool definition and the attachment allowance both come from proposed regulations. When the final text is confirmed, update the pool test and the link rule.

When the standard changes, re-grade every live sequence against the new version rather than waiting for the next search. A Cleared grade is only valid against the version of the standard it was graded under, so record that version with every grade.

#### Before any search goes live

- [ ] Every candidate location is tagged against the New York reports-to test and the Colorado could-work-there test.
- [ ] The employer's headcount has been checked against each triggered jurisdiction's threshold.
- [ ] A dated range record with a written rationale exists for every location and level in the search.
- [ ] No range is open-ended, and the max-to-min ratio has been computed and reviewed for every range.
- [ ] Every template and sequence step is labelled pool or one-to-one by recipient count per template.
- [ ] Every pool message carries the range in the body, not only behind a link.
- [ ] Colorado, Washington, New Jersey, Illinois, and Maryland reach includes bonus and benefits wording, and Colorado reach includes an application deadline.
- [ ] New York State reach includes a job description, and the NYS and NYC range is base pay only.
- [ ] Every screen script has been searched for "current" and "previous" and contains expectations questions only.
- [ ] The actual agency and AI-drafted scripts for this search have been collected and graded.
- [ ] Two reviewers graded each item independently and recorded the same grade with the standard version.
- [ ] Owners and dates are set for the Washington 5-business-day and NYC 30-day cure windows.

## Frequently asked questions

### Do recruiter messages need a salary range?

If the same template goes to more than one person, treat it as a job advertisement and include the range. New York State's proposed regulations count an ad sent to a pool of more than one prospective applicant, and they name e-mails to a pool and mailing-list messages. No regulator has said whether a truly one-to-one message is exempt, so the safe default is to include the range whenever a message names a specific open role in a covered jurisdiction.

### Can recruiters ask current salary on a screening call?

Not in a script you want graded Cleared. Around 20 states plus DC ban pay-history questions for private employers, and the soft version is still banned, because asking whether the candidate would like to share current pay counts as prompting. Ask what the candidate expects to earn in the next role. Illinois, Colorado, and Kansas City expressly allow that question, and it is the only compensation question this standard keeps.

### Which state's pay transparency rule applies to a remote role?

Every state the search reaches can apply, and the strictest one decides the grade. New York covers a role if the person reports to a supervisor, office, or work site in New York. Colorado covers a remote posting even if it says Colorado applicants will not be accepted. A single remote sequence can trigger both tests, so it needs both sets of disclosures.

### Am I responsible for what an agency recruiter or AI tool asks candidates?

In several jurisdictions, yes. California bans pay-history questions asked personally or through an agent, Maryland bans them through an employee or agent, and Connecticut bans them directly or through a third party. NYC covers employment agencies of any size. Collect the actual scripts and drafts that agencies and AI tools use for each search, and grade them the same way you grade your own.

### Is a link to the posting enough to disclose the range in an email?

In Colorado, yes, as long as the range is hyperlinked or reachable through a URL in the message. Whether New York's similar allowance made it into the final rule is unverified. A link only counts while it works. If the posting is taken down or edited, the message no longer carries a range, so this standard requires the range written in the message body as well.

### Does the EU Pay Transparency Directive apply to recruiter outreach?

It sets a minimum, and national laws decide the details. Article 5(1) requires the starting pay or range to be given in a published vacancy notice, before the interview, or otherwise, and Article 5(2) bans pay-history questions. Italy, Lithuania, and Slovakia require the range in the advert itself. How the national laws treat intermediaries is not publicly established, so put the range in outreach to candidates in those countries.

---

*From the Refolk guide library. I revise these guides rather than replacing them, so the current version is always at https://www.refolk.ai/guides/outreach-pay-disclosure-standard*