OpenAI audio uploads and Greenhouse Voice AI blur interview prep and screening

What changed and why it matters

If you are preparing for an interview or an exam, this week’s product updates matter because the same general class of AI tool is moving in two directions at once. According to OpenAI, paid ChatGPT users can now upload audio for transcripts, summaries, and questions about recordings. At the same time, Greenhouse is pushing its Voice AI product deeper into hiring, making live screening more software-driven.

That overlap creates a clearer link between what candidates use to rehearse and what employers may use to evaluate them. OpenAI’s release notes say a meeting, interview, or lecture can be turned into structured notes or a follow-up email draft. Greenhouse’s materials, meanwhile, describe a two-way voice interview built around the questions and scoring criteria a hiring team defines.

OpenAI’s audio uploads make preparation more structured

On October 6, 2026, OpenAI added audio uploads to ChatGPT for transcripts, summaries, and questions about recordings. The feature is available with paid ChatGPT subscriptions and workspaces, though OpenAI says availability may vary by workspace settings, region, client version, and model.

For someone practicing for an interview or exam, the practical use is straightforward: record a rehearsal, turn it into notes, and use the result to spot weak answers or confusing sections. OpenAI explicitly says users can turn a meeting, interview, or lecture into structured notes or a follow-up email draft. That makes the tool more useful for preparation, but also more dependent on the quality of the recording and the transcription.

OpenAI also warns that transcripts may contain errors and that performance may vary across languages. For users, that means the output is helpful for organizing thoughts, but it should not be treated as perfect or authoritative without review.

Greenhouse is building live voice screening into the hiring stack

On October 15, 2026, Greenhouse published a post arguing that hiring is trapped in an AI-driven doom loop of bots, spam, and lost trust. Its proposed remedy is an orchestrated hiring ecosystem where interviewing, sourcing, scheduling, and the ATS work together. In the same period, Greenhouse’s newsroom said it had completed its acquisition of Ezra AI Labs, the voice AI interviewer, following a deal first announced on May 5, 2026.

Greenhouse says interviews open up to more candidates through Voice AI and that scheduling happens when a role opens rather than days later. Its Voice AI product is a two-way voice interview that uses the questions and scoring criteria a hiring team defines. Greenhouse says the product is currently available for purchase in the US and internationally.

The company also says its Ezra AI interviewer now supports 10 additional languages, including Spanish, Portuguese, German, French, Italian, Dutch, Russian, Hindi, Japanese, and both LATAM and Spain variants of Spanish. Greenhouse says independent bias audits are conducted every month by Warden AI and published publicly. On its product page, Greenhouse says Voice AI is intended to keep every decision human, with every score overridable.

The rules matter more as the tools get closer together

The practical question for candidates is not whether AI exists in interview prep or hiring, but where the organization draws the line. OpenAI’s features are positioned around note-taking, summaries, and follow-up drafting. Greenhouse’s product is positioned around live interviewing and screening. Those are different uses, but they now sit much closer together in the workflow.

That makes it more important to read the employer’s or school’s rules before using any assistant in a live process. For preparation, practice beforehand and use the tool to review recordings, organize answers, and identify gaps. If a policy requires disclosure, disclose the tool. If an automated screen seems wrong, ask for review through the organization’s process rather than trying to work around it.

The broader pattern is that interview-adjacent AI is becoming more capable and more ambiguous at the same time. OpenAI and Greenhouse are moving fast in parallel: one side is aimed at prep and note-taking, the other at live screening. For candidates, that means the boundary between rehearsal software and evaluation software is getting harder to ignore.

What to watch next

The most useful next signals will come from product pages, release notes, and employer policies. According to OpenAI, availability for audio uploads may vary by workspace and region, so the details matter. According to Greenhouse, the company is emphasizing multilingual access, public bias audits, and human override controls, which are all relevant to how such tools are adopted.

For now, the safest approach is simple. Use assistants to prepare before the test or interview, check the published rules, and follow the process if something automated appears to be inaccurate. As these products continue to merge preparation and screening, the key issue is not just capability, but clarity about when and how the tools are supposed to be used.