How to Vet a Nanny and Household Staff: A 2026 US Checklist

Vetting people is my job, and "please check our nanny" reaches us more often than any other family request. It almost always sounds the same: we found someone through an agency or an app, the platform says she passed a background check, but we would like to be sure. And it almost always turns out the family doesn't know what that check covered — not out of carelessness, but because nobody told them a background check has edges.

The articles online do not help much. The top-ranking pieces on vetting a nanny tell you to "run a background check" and list interview questions. None of them explains what a standard report misses, or which free public sources you can check yourself in twenty minutes. This article does, and that is the part worth taking away.

Three levels of vetting: what you do yourself, what to buy, what nobody should sell you#

Before opening any database, it helps to set the frame. A nanny, a driver and a housekeeper form a trust perimeter — the circle of people with physical access to your home and knowledge of the family's routine. An insider inside that perimeter is more dangerous than any outside attacker for one reason: they do not need to break anything. The door has already been opened for them.

Hence three levels of work:

1. Do it yourself, mostly free. Documents and their authenticity, public records, calling referees, the open digital footprint. This is most of the practical value.
2. Worth paying for. An FCRA-compliant background check with county-level searches; a structured interview with cross-verification of the candidate's account; identifying referees; correlating the digital footprint across several identifiers.
3. Nobody should sell you this. A candidate's phone records or bank details, "reports" built from leaked data, covert audio recording. Parts of this are federal crimes — more on that in the section on the legal perimeter.

There is a fourth item nobody can give you: a guarantee. Vetting lowers probability; it does not remove risk. Anyone promising otherwise is selling reassurance, not security.

Documents: don't take them on trust, verify them#

The list is much the same everywhere — ID, work authorization, a background check, certifications. What almost nobody explains is where each one stops. That is the actual work.

Identity and work authorization: Form I-9#

If you employ a nanny directly, you are an employer, and a household employer must complete Form I-9 for every employee hired to work in the US — part-time and live-in included. The exceptions are narrow: sporadic, irregular work such as an occasional babysitter, and workers supplied by an agency that does the I-9 itself.

Three details that decide whether the form means anything:

• Look at the original documents, in person. The candidate fills in Section 1 by the first day of work; you examine the documents and complete Section 2 within three business days of the start date. A photo of a card sent by text is not an examination.
• Let the candidate choose the documents. The form lists acceptable combinations, and asking for a specific one ("bring your green card") is a form of discrimination the law prohibits.
• E-Verify is optional for household employers unless your state requires it. It checks the I-9 data against government records, but only after the form is done.

Keep the form: three years after the hire date or one year after employment ends, whichever is later. And the money is not the point of doing this properly: a person without status working in your home is exposed to pressure from anyone who knows about it.

The background check: what a report shows and what it leaves out#

There is no single national criminal record a private family can pull. Criminal cases live mostly in county courts, and a background check is only as good as the list of places it searches. A typical report from a consumer reporting agency (CRA) combines:

• an SSN trace — not a verification of identity, but a list of addresses linked to the number in credit-header data. Its real job is to tell you which counties to search;
• county criminal searches in those counties — the most reliable part;
• a "national" database search — an aggregate of whatever courts and agencies sell to data vendors. Useful as a pointer, incomplete by design;
• the sex offender registries;
• optionally a driving record (MVR) and, where your state and city allow it, a credit report.

Now the part this section exists for. What the report does not show:

• counties nobody searched. If the candidate lived in five counties and the package covers the current one plus a database, the other four are a blind spot;
• anything that never became a case: someone quietly let go for theft, no police report — the report will be clean;
• sealed and expunged records, and, under the FCRA, arrests that did not lead to a conviction once they are more than seven years old. Several states, including California and New York, also limit how far back convictions can be reported;
• records from outside the US. For staff who have lived abroad, some CRAs run international checks, but coverage varies by country;
• civil disputes and debts, unless you order those searches separately.

Hence the logic of everything that follows: the report is a necessary but nowhere near sufficient filter, and the next steps cover precisely its blind spots.

The process is part of the law. A background check ordered through a CRA is a consumer report under the Fair Credit Reporting Act, and the FCRA applies to you as a household employer. Before ordering, give the candidate a standalone written disclosure and get their written authorization. If you decide not to hire because of something in the report, send a pre-adverse action notice with a copy of the report and the summary of rights, wait a few business days so they can dispute errors, then send the final adverse action notice. Some states and cities add fair-chance rules on top — New York City, for example, only allows criminal history questions after a conditional offer — so check whether yours covers household employers.

The correct wording in negotiation is not "I am entitled to demand this" but "in our family this is a condition of employment." A refusal to authorize a check, from someone willing to discuss everything else, is information in itself.

CPR and first aid: verified by ID in thirty seconds#

A certificate is the one credential almost every family asks for and almost none verifies. Both major providers make it trivial:

1. American Red Cross — enter the certificate ID or scan the QR code at redcross.org/take-a-class/digital-certificate.
2. American Heart Association — enter the eCard code or scan its QR code at heart.org/cpr/mycards.

The practical conclusion is simple: ask not for "the card" but for the card and a scan in front of you. A candidate who "has it somewhere at home," and a candidate whose certificate does not come up, are two different conversations — and the second one ends there.

One caveat. A certificate proves someone sat through the course on a date. It says nothing about whether they would keep their head with a choking toddler — that is what the structured interview and the references are for.

Education and licenses#

Degrees are confirmed through the school's registrar or, for a fee, through the National Student Clearinghouse. For licensed roles the state does the work for you: nurses through Nursys, nursing assistants and home health aides through the state registry, teachers through the state education department.

A record that doesn't come up is not a verdict but a prompt for a follow-up question: where and when did you study, what name were you licensed under? A genuine graduate answers without hesitating.

Free public records: twenty minutes, zero cost#

This is the part missing from almost every page in the search results. All the sources below are public and work from a browser.

State child abuse and neglect registries are the obvious next question, and the honest answer is that it depends on the state. In some, a candidate can request a clearance on themselves and hand it to you; in others access is limited to licensed childcare providers, and some states forbid other employers from requiring it. A reputable screening company or your state's child welfare agency will tell you what is possible where you live.

And the main idea to hold on to while working through records. They do not answer "is this a good person"; they answer "does what I was told match what is documented." An old bankruptcy is no reason to refuse: plenty of honest people have been through one. A fresh judgment alongside a claim of "no financial problems" is a reason — because the problem is not the debt but the willingness to lie to an employer about money while living in their house.

References: telling a real employer from a friend#

"Call the previous employers" is advice everyone gives. The trouble is that a dishonest candidate brings along a briefed acquaintance who will say exactly what the reference letter says.

The working technique is to test memory, not the story. A fake referee will reproduce the letter, but will not survive the details that are not in it.

1. Identify the referee rather than accept a contact. A number handed to you by the candidate proves nothing. Look at whose number it is: name and photo on messaging apps, linked accounts, whether the person exists in public records and on LinkedIn at all.
2. Ask what is not in the résumé. How old were the children when the nanny left, what were their names, which neighborhood did the family live in, how did she get there, who else worked in the house.
3. Build the chronology backwards: "when did it end" → "when did it start" → "what happened between that job and the next one." Gaps of three to six months are the most common point where a story falls apart.
4. The red flag is not criticism but sterility. "Everything was wonderful, no complaints at all," with no single concrete situation, is the classic fake-referee answer. A real employer almost always recalls at least one rough patch: lateness, a disagreement about routine, something ordinary.

The candidate's digital footprint: what to look at and what it means#

The advice to "check their social media" also travels from article to article, and everywhere it takes exactly one line. The work, though, is not to "read a profile" but to connect identifiers: phone number → messaging accounts → username → other platforms → photographs. Each step gives you an independent confirmation or contradiction of the account you were given.

The order of work:

1. From the number. Which messaging apps are linked to it, what name and photo are on display, how the number is labelled by other people — caller-ID apps such as Truecaller show those crowd-sourced labels, and past employers and agencies that never made the résumé tend to surface there. A caveat: these apps build their data from their users' uploaded address books, and that is worth understanding while using them. More on what a number exposes is in a separate piece on finding someone by phone number.
2. From the username. Enumeration across platforms. The main difficulty is false positives on common handles, so every hit is confirmed by hand against the avatar, registration date and writing style. Tooling and method are in the article on finding someone by username.
3. From photographs. Reverse image search on the profile picture surfaces accounts opened under another name; how that works is covered in the piece on finding someone by photo.
4. Correlation with the account given. Do the cities and periods match the story about previous jobs — and the address history in the background report?

The red flags here are not the ones amateurs look for. It is not about "bad photographs":

• a gap between the story and the footprint — "five years with one family" alongside regular posts from another state during the same period;
• traces of a debt spiral — public requests to borrow money, payday-loan groups; combined with a recent judgment in the court records this stops being a guess and becomes a picture;
• demonstrative publicity about employers — photographs of other people's children and other people's interiors in an open profile. This is the single best predictor available, and the logic is direct: she will display your home the same way;
• other people's faces in "her own" photographs — borrowed material in the profile.

The other direction: what your staff knows and posts about you#

What follows is the most underrated part of the subject. Every article on vetting a nanny looks one way: check the candidate. But a nanny and a driver know things about your family that no outside attacker does — what time the child is collected from school and by which route, who is home during the day, when the family travels, what the house looks like and what is in it.

This material leaks not through malice but through everyday publicity.

Leak channels#

• Geotags and real-time posts. "We're at the playground" plus a location tag equals an address and a time, every weekday.
• Photographs of the child in a nanny's open profile: the child's face becomes searchable by face-recognition tools.
• Photographs of the interior — a shop window for your possessions.
• Nanny groups and neighborhood chats — by town, by building, by agency. Employers, pay and routines get discussed there. It is an environment with no access control at all.
• A personal phone in the house — cloud backups and photo auto-upload into someone else's accounts.
• Smart home and cameras, to which staff were given access "to make things easier."

What the contract and device rules cover#

1. A confidentiality clause in the work agreement, or a separate NDA: no publishing photographs of the children, the house, or any information about the family; no geotags; consequences for a breach.
2. A device rule: a personal phone, yes; filming the child on a personal phone, no; work photographs only into an agreed family chat.
3. Separation of access: a separate smart-home account, a separate door code, a separate guest Wi-Fi. All of it revoked on the day employment ends — that day, not "some time this week."
4. A periodic audit of your own footprint — what is already visible about your family from outside. The reasoning and the tooling are in the pieces on OPSEC for executives and on removing your personal information from the internet.

The polygraph: what it actually shows#

Search for household-staff vetting and a good share of the results are polygraph examiners. Their claim is uniform: the test will reveal cruelty to children, theft and substance problems, and a refusal to take it means the person has something to hide. That claim deserves a proper examination, because it is half wrong.

What the science says. The reference work here is the National Research Council report The Polygraph and Lie Detection (2003). Its findings:

• for specific incidents, the comparison test distinguishes lies from truth well above chance, but well below perfection; median accuracy is around 85 %;
• in screening — and hiring a nanny is screening — the method breaks against the base rate. In a group where genuine wrongdoers are rare, the test produces an avalanche of false positives. The report's own illustration: among 10,000 employees with 10 actual wrongdoers, tuning the test to catch about 80 % of them would see roughly 1,600 people fail it;
• accuracy degrades under countermeasures — and precisely in the motivated wrongdoer, the one person the test exists for.

Apply that to your situation. You are testing one person rather than ten thousand, but the arithmetic is identical: the base rate of "nanny who is a criminal" is vanishingly small, so the overwhelming majority of "suspicious reactions" in screening are false alarms on honest candidates. At the hiring stage a polygraph filters out the anxious, not the dangerous.

What the law says. The federal Employee Polygraph Protection Act bars most private employers from polygraph screening of applicants. Whether it reaches a family employing one nanny is not settled, and several states restrict polygraphs in employment on their own — so talk to an employment lawyer before making a test a condition of hire. And in court the result is of little use: most US courts exclude polygraph evidence or admit it only when both sides agree in advance.

The conclusion I give clients. A polygraph is not a hiring filter but, at most, a tool for examining a specific incident: a sum has gone missing, an item was damaged, the circle of people is defined. There the base rate is high and the method performs at its best. As for "she refused, so she's guilty," it is worth naming plainly: declining a procedure that is 85 % accurate and rarely admissible is rational adult behaviour, not a confession.

Nanny cams: where protection ends and a crime begins#

"Put in a camera" is advice given casually. In reality there are three separate legal questions here that usually get collapsed into one.

Video. Recording silent video in the common areas of your own home is legal in most states, whether or not the camera is visible. The hard line is places where a person has a reasonable expectation of privacy: bathrooms, and a live-in nanny's bedroom. A camera there turns you from the party protecting a child into the defendant — many states treat it as a crime.

Audio is a much stricter regime, and this is where families get caught. Federal wiretap law lets you record a conversation you are part of. A nanny cam recording the nanny on the phone, or talking to a friend at the playground while you are at work, is recording a conversation you are not part of — without anyone's consent. On top of that, about a dozen states, including California, Florida, Illinois, Maryland, Massachusetts, Pennsylvania and Washington, require the consent of every party to a recording. The simple rule: audio off, unless the nanny has agreed to it in writing.

Disclosure is the workable option. The work agreement records where the cameras are, whether audio is captured, how long footage is kept and who has access. No cameras in bathrooms or in the nanny's own room.

Hence the conclusion that inverts the casual advice: a disclosed camera protects you better than a hidden one. It deters, it keeps you clear of the audio trap, and if you ever need the footage — for a police report or a lawsuit — nobody argues about how it came to be recorded.

And a layer nobody raises in the nanny context: a home camera is itself a leak vector. Default passwords, a stream exposed to the internet, a contractor's cloud access — a surveillance setup installed for a child's safety can, if configured badly, broadcast the family's daily routine outward. Whether your own credentials have surfaced can be checked against breach databases.

Roles: driver, housekeeper, security, carer#

The common package is the same for everyone, but each role carries its own risk profile.

A personal driver. Two risks a nanny does not carry: he operates a car with your children inside it, and he knows every route and every schedule. In addition to the standard checks:

• the driving record (MVR) — suspensions, violations, accidents. Order it as part of the background check with the candidate's written authorization, or ask the candidate to request their own record from the state DMV and hand it to you. Under the Driver's Privacy Protection Act you cannot simply pull someone else's record without consent;
• license status and insurance — whose policy covers the car, and whether the insurer knows a hired driver uses it;
• route discipline as a separate clause: no publishing locations, no discussing the family's schedule, rules for personal phone use in the car.

A housekeeper has maximum access to property and documents with minimal observation. The key here is not background but regime: a list of access zones, a procedure for handling cash and purchases, separate storage for documents and valuables.

A carer has access to medication, bank cards, and a vulnerable person who may not complain. Licensed roles — nursing assistant, home health aide, nurse — are checked separately through the state registry or Nursys.

Security is the one role where skipping the check is indefensible: this person knows your security arrangements from the inside. In most states armed and unarmed guards are licensed, and the license is searchable.

Paperwork: employee status, nanny taxes and the work agreement#

A nanny you hire directly is almost always your employee, not an independent contractor: you decide what she does and how, and that is what the IRS looks at. Calling the arrangement a contract does not change it. The consequences, per IRS Publication 926:

• Social Security and Medicare apply once you pay a household employee $3,000 or more in cash wages in 2026 ($2,800 in 2025). You withhold the employee's share and pay your own.
• Federal unemployment tax applies if you pay $1,000 or more in total to household employees in any calendar quarter.
• You report it all on Schedule H with your own tax return and give the employee a W-2.
• State rules come on top: unemployment insurance, and in some states workers' compensation and paid leave. They vary widely.

Federal wage law applies too: domestic workers are entitled to minimum wage, and live-out nannies to overtime; live-in nannies are exempt from federal overtime, though some states cover them.

A written work agreement is where the vetting pays off: duties, hours, pay, the confidentiality clause, the camera and device rules, and what happens on termination.

Cash in hand, off the books. The core point: without paperwork you have no documented employment relationship at all. Recovering damages, proving what was agreed, calmly reporting a theft — all of it gets harder, and the tax question can surface at the worst possible moment. That is not hypothetical: unpaid taxes on household workers ended a US Attorney General nomination in 1993, and the episode still has a name — Nannygate.

The legal perimeter of vetting: what you must not buy#

This is the one section I consider required reading even if you skim the rest.

Pretexting is a federal crime. Obtaining someone's phone records by impersonating them or otherwise lying to the carrier is punishable by up to ten years in prison under the Telephone Records and Privacy Protection Act of 2006. Obtaining financial information from a bank the same way is prohibited by the Gramm-Leach-Bliley Act. So anyone offering a candidate's call log, text history or bank balances is describing a crime — and you would be the customer.

People-search sites are not background checks. They say they are not consumer reporting agencies under the FCRA, and their terms forbid using the results for hiring. The data is aggregated from data brokers without the accuracy duties a CRA has, and the person it is wrong about gets none of the FCRA dispute rights. For a hiring decision, use a CRA and follow the FCRA steps above.

Recording laws apply in your own home. Covered in the section on cameras: audio without consent is the fastest way for a careful parent to commit an offense.

Private investigators are licensed in most states and work under the same laws as everyone else. They have no special access to criminal records beyond what a CRA can search, and an investigator promising phone records or bank statements is describing pretexting, not skill.

What makes vetting lawful: the candidate's written authorization for the specific checks, obtained the way the FCRA requires. That is a protective document for you, not a formality for them.

To put the whole subject in one sentence: vetting a candidate is measured not by how much you can dig up but by how much you can learn without creating a problem of your own. The lawful perimeter is documents produced voluntarily, public records, the open digital footprint, and a background check with consent. Everything beyond it is bought from people whose trade is itself illegal — and those people arrive in your home along with the report.

Vetting isn't a one-off: when to re-check#

A check records a state on a date, and risk moves. Someone impeccable three years ago may since have taken on debt, developed a dependency, or come under pressure from third parties. Corporate security handles this through continuous evaluation — periodically revisiting clearance instead of vetting once at hire; both RAND and CISA have examined the approaches in their insider-risk guidance. Transferring that model to a household is our own interpretation, but the logic carries over exactly.

The household version is three repeatable actions and a list of triggers:

• every 6–12 months: re-run the free searches — the sex offender website and the court portals. A new case during employment means far more than a clean record at hire. Re-running a paid background check needs the employee's authorization under the FCRA, so get it in writing;
• annually: the driving record for a driver, and certificate expiry dates for CPR and first aid;
• on a change of circumstances: divorce, a move, a sharply changed spending level, requests for loans, a new circle of acquaintances, the appearance of "a friend who'll cover sometimes."

I press that last point separately. A stand-in is the most common way a completely unvetted person gets into a house.

If the check turns something up: parting ways#

Every article ends at "you've hired someone." But a person refused after a check knows more about your home than a random candidate, and that is a task of its own.

The principles we apply in practice:

• if the decision rests on a background report, follow the FCRA adverse action steps — that is the one situation where you must tell the candidate where the information came from;
• otherwise, do not say what you found or where. It protects the source and avoids provoking a confrontation;
• decline without a detailed rationale. Accusations help nothing here;
• revoke every access on the day you part — codes, keys, smart-home accounts, guest Wi-Fi, family chats. Not "some time this week";
• close out pay and paperwork in full. Money owed to a former employee is the most common trigger for a conflict that then travels well beyond money.

What a paid check buys you#

An honest breakdown, because nobody selling these services provides one.

Free and within your reach: the I-9 document review, the sex offender website, county court portals, PACER, license and certificate lookups, calling referees, the open digital footprint.

Cheap to buy: a database-only "national" report — it saves time, but it misses exactly the counties and cases that matter. If a package costs less than a single county court search, a database lookup is what it is.

Worth buying: an FCRA-compliant check with county searches across the candidate's full address history, plus an MVR for anyone who will drive your children.

Requires a specialist: a structured interview with cross-verification of the candidate's account; identifying referees rather than taking their contacts on trust; correlating the digital footprint across several identifiers; examining a specific incident once one has occurred.

If you need to vet not one person but the whole perimeter — nanny, driver, housekeeper, security — and to do it regularly, it makes sense to treat professional staff vetting as a process rather than a one-off service. But start with the free minimum in this article regardless: it closes much of the risk and costs you an evening.

FAQ#

How do I run a background check on a nanny?#

Order it from a consumer reporting agency, not a people-search site. Give the candidate a standalone written disclosure and get written authorization first, as the FCRA requires. Make sure the package includes county criminal searches for every county in the candidate's address history, not just a "national" database. Search nsopw.gov yourself — it is free.

What should I verify besides a criminal record?#

Identity and work authorization on Form I-9, with the original documents in front of you. A CPR or first-aid certificate by its ID on the Red Cross or AHA site. References, by identifying the referee rather than taking the contact on trust. For a driver, the driving record.

Is it legal to have a nanny cam?#

Silent video in the common areas of your home is legal in most states. Never place cameras in a bathroom or in a live-in nanny's bedroom. Audio is the risk: recording conversations you are not part of can break federal and state wiretap laws, and about a dozen states require every party's consent. Disclose the cameras in the work agreement and keep audio off unless the nanny agrees in writing.

Do I have to pay taxes for a nanny?#

If you pay a household employee $3,000 or more in cash wages in 2026, you owe Social Security and Medicare taxes, and federal unemployment tax applies once you pay $1,000 in a quarter. You report them on Schedule H and give the nanny a W-2. State taxes and insurance requirements vary.

Is a polygraph worth using on a nanny?#

As a hiring filter, almost useless: screening a group with a low proportion of wrongdoers generates a mass of false positives, the federal polygraph law restricts it for most employers, and courts rarely admit the results. A polygraph is at most suited to examining a specific incident with a defined circle of people.

How do I verify a nanny's references?#

Identify the referee rather than taking the contact on trust, and ask for details absent from the reference letter: the children's names and ages when she left, the neighborhood, the route, who else worked in the house. A fake referee will reproduce the letter but will not survive the details.