How to get AI interview notes into an ATS without an integration

If your ATS has no native connection to your AI notetaker, you have four routes: paste the note by hand, email it to an address the ATS logs, connect the two with Zapier, or write to the ATS API. As of October 2026, Zapier and the API are both workable for notes: Greenhouse documents a Zapier "Create Candidate Note" action, and the Bullhorn, Greenhouse and Lever APIs all document creating notes. Email is the weakest route, and automated scoring is the riskiest. For one recruiter or a small desk, pasting a well-shaped note usually wins on cost and control.

This page compares the routes, with the ATS facts taken from each vendor's own help center or developer documentation, and ends with what to check before you automate anything. If your question is which notetakers already integrate with your ATS, start with our guides for Greenhouse, Bullhorn, Lever and JobAdder.

Which route should you use?

Is copy and paste good enough?

Work it out before you build anything. Multiply the minutes a paste takes, including opening the record and fixing the format, by your calls per week. As an example only: two minutes a paste across 25 screens a week is under an hour a week per recruiter. If an integration project would take a developer several days plus upkeep, pasting is cheaper for a long time, and it keeps a person looking at every note before it becomes part of the record.

Pasting works well when three things are true:

• The note is shaped for the destination. Bullhorn notes are plain text only, per its knowledge base, so Markdown arrives as stray symbols. A Greenhouse scorecard wants evidence per question, not one summary block. Ask your notetaker for the right shape, or keep a template per ATS. Some tools do this for you: Interview Signal, for instance, has no ATS integration, but its export copies a note shaped for Bullhorn, Greenhouse, Lever or JobAdder to paste in.
• The note is short. Verdict, evidence against each requirement, logistics, next step. Transcripts belong wherever your retention policy puts them, not in a note field.
• Everyone pastes into the same place. Agree which note type or scorecard field each kind of call goes in. Our ATS-specific pages cover this for Bullhorn, Greenhouse and Lever.

Can you email AI notes into the ATS?

Only where the ATS documents it, and even then email is built for correspondence, not notes. Greenhouse is the clearest example. Its Maildrop article (updated March 2, 2026) describes a general Maildrop address for logging email conversations with candidates, by forwarding a thread or BCCing the address, plus a unique address per open job for creating candidates. Four details make it a poor route for interview notes:

• Maildrop only accepts mail from an address linked to your Greenhouse user account, and auto-forwarding from your email client is not supported.
• It ignores addresses belonging to internal users, by design.
• If a message without a resume goes to the general address, Maildrop tries to build a candidate profile from the email itself, and creates a new prospect if the candidate's email is not found. A notes email from your notetaker could create a stray prospect rather than a note.
• What it logs is an email in the activity feed, not a candidate note with note visibility.

Lever's help center describes two-way email sync for messages your team sends to candidates through Gmail or Microsoft 365; we found no documented address for posting an internal note by email. Whatever your ATS, confirm in its own documentation that an address creates a note on an existing candidate before you rely on it.

Is Zapier workable?

Yes, if two things exist: a trigger your notetaker can fire (a new email, a new file in a cloud folder, a webhook or a new spreadsheet row) and an action in your ATS's Zapier app that creates a note. Desktop notetakers often have no webhook, so check the trigger side first.

• Greenhouse. Greenhouse's Zapier integration article (updated July 9, 2026) lists a Find Candidate search by email address and Create Candidate, Create Candidate Note and Create Prospect actions. It notes that Greenhouse is a premium Zapier app needing at least a Starter plan, that the integration works for up to 100 jobs, and that setup uses a Harvest API key. Greenhouse's Harvest documentation says Harvest v1/v2 is deprecated with removal on August 31, 2026, so test that your zap still writes notes and ask Greenhouse support which key type it now uses.
• JobAdder. Zapier's JobAdder app page lists Add Candidate Note and Add Application Note actions, each needing the record's ID, a note type and the note text, as of October 2026.
• Bullhorn and Lever. Check Zapier's app directory and your ATS's marketplace for a current app with a note action before planning around one; if there is none, the API is the route.

Build the zap as find, then write, then alert: search for the candidate by email, create the note only if exactly one candidate matches, and send yourself a message when none or several do. Never let the zap create a candidate to have somewhere to put a note. Duplicate records are harder to fix than a missing note.

Is the ATS API workable?

For teams with a developer, yes. All three ATSs below document creating notes; Greenhouse and Lever also document creating feedback or scorecards. As of October 2026:

Two details from these docs shape the build. Greenhouse's Create note says interview and feedback notes are produced by Greenhouse features and cannot be created through that endpoint, so interview feedback either goes through the scorecard endpoint or stays as a plain candidate note. And Bullhorn's Note entity links the note to the person automatically when you send both commentingPerson and personReference; leave either out and you have to create the association yourself.

What should you check before you automate?

1. Matching. How is the candidate found: an ATS ID, an email, a name? What happens with no match, two matches, or a candidate who used a personal email on the call and a work email on the application?
2. The object. Candidate, application, contact or job. In Greenhouse a note can be anchored to one application; in Bullhorn a note written against a job is associated with the contact, not a candidate.
3. The author. Bullhorn's commentingPerson defaults to the user creating the note, Lever writes on behalf of a perform_as user, and Greenhouse acts as an integration user unless the token names a Site Admin. A feed full of notes "by" an API user is hard to trust.
4. Visibility. Decide what is private before the first note: pay, competing offers and health details should never land in a note the whole hiring team reads.
5. Format and length. Plain text in Bullhorn; Greenhouse's API stores line breaks as written. Send the summary, not the transcript.
6. Consent. The candidate agreed to be recorded for this interview; check your notice covers storing the AI summary in the ATS. Our consent script has a "where it goes" line for exactly this.
7. Deletion. When a candidate asks to be deleted, list every place a copy now lives: the notetaker, the automation tool's logs, the ATS.
8. Credentials. Who owns the key, what it can read, and what happens when that person leaves. Greenhouse's own support pages note that candidate notes, including private ones, are readable by anyone holding an API key with activity feed permission.
9. Failure alerts. An automation that fails silently is worse than pasting, because everyone assumes the notes are there.

Should AI write scores into the ATS automatically?

Not without a person in between. A note is context; a score moves a candidate. If you use Greenhouse's scorecard endpoint, create drafts and let the interviewer submit, because a scorecard created as complete is submitted and locked outside your organization's edit window. In Lever, the four-point rating on a feedback form is the interviewer's call, not the model's. The safest automation writes the evidence and leaves the judgment, and the submit button, to the person who ran the interview.

A sensible order: paste for a month and agree the format; automate plain notes for one team; measure how often the match fails; and only then consider drafts of feedback or scorecards.