Disability & Policy · Article
SiriusXM and Pandora Agree to Real Podcast Transcripts for Deaf Listeners, Ending a Five-Year ADA Fight
A proposed consent decree requires named speakers, sound descriptions and human-edited transcripts for 30 top shows, and ties accuracy to named speech-recognition tools. It takes effect when a federal judge signs it.
Published · 7 min read
By Kaeleb Savon Heck
DAVAO DEL NORTE, Nov. 19, 2026 — SiriusXM and Pandora have agreed to provide full, readable transcripts for the podcasts on their streaming services, with speaker names, descriptions of music and sound effects, and real punctuation, under a proposed settlement that would end a five-year disability rights lawsuit brought by Deaf and hard-of-hearing listeners.
The consent decree was filed on Nov. 16 in the U.S. District Court for the Southern District of New York. It would resolve a 2021 case brought by the National Association of the Deaf (NAD) and four Deaf and hard-of-hearing individuals, who argued that the companies broke the Americans with Disabilities Act and New York law by offering podcasts with no transcripts at all. The deal takes effect once Judge Jeannette A. Vargas signs it, and it would last for three years.
Beyond the promise of transcripts, the agreement is notable for how specific it is. It names the speech-recognition tools the companies must match, sets out which shows get human editing, requires a "Report a problem with the transcript" link, and gives the plaintiffs a role in writing the style guide. Advocates say it could become a benchmark for an industry where transcripts are still often missing or barely readable.
What the settlement requires
According to Disability Rights Advocates (DRA), the nonprofit law firm that represented the plaintiffs, transcripts for podcasts that SiriusXM and Pandora own or license must include:
- Speaker identification.
- "Atmospherics," meaning non-dialogue audio such as sound effects, laughter and music.
- Proper punctuation, paragraph breaks and capitalization.
The decree itself defines speaker identification as using a speaker's actual name, such as "Bob" or "Janet," wherever possible, rather than generic labels like "Speaker 1." When a name is not known, the transcript should use a descriptive label such as "Interviewee" or "Passerby."
The companies must also use "human cleanup," a person reviewing and editing a machine-made transcript, on 30 podcasts that the plaintiffs will choose each year from the companies' top 50 revenue-generating shows. The decree lets the companies later propose ending human cleanup if speech-recognition quality improves enough, but the plaintiffs can object and the dispute process applies.
Other commitments listed by DRA and in the decree include custom data dictionaries of words and phrases used on specific shows, an annual review of transcript quality and of the best available transcription vendor, a style guide developed with the plaintiffs' feedback, and annual staff training on Deaf culture and communication access. The decree says that training will be delivered remotely through Coursera and led at least in part by a Deaf organization or a Deaf instructor.
The fine print: named tools and accuracy floors
The consent decree is unusually concrete about technology. It requires the companies to transcribe new episodes of all podcasts they own or license using ElevenLabs Scribe v2, and later a "Transcript Enhancement Service" that the companies must try to build within six months of the effective date. That service is meant to clean up machine transcripts, tell speakers apart, label them and apply consistent formatting.
For third-party podcasts, shows licensed under standard agreements that are not individually negotiated, the decree requires transcription using Rev.
In both cases, the accuracy standard is tied to those tools as of the effective date. If the companies switch vendors, they must keep accuracy, speaker identification, atmospherics and formatting at least as good, and tell the plaintiffs about the change. Every transcript must also say on its face how it was made, for example by automatic speech recognition alone or with human edits.
The annual quality evaluations must look at word error rate, grammar, speaker-identification accuracy, how sound descriptions are labeled and placed, and whether formatting follows the style guide. The companies must keep records of complaints and report the numbers, the nature of the problems and how they were resolved to the plaintiffs every year. They must also post a public accessibility statement and give the plaintiffs guest accounts to monitor the services.
Money and enforcement
The decree requires the companies to pay $5,000 in compensatory damages to each of the four individual plaintiffs and $15,000 to the NAD on behalf of its members. They will also pay $1,465,000 to Disability Rights Advocates for attorneys' fees, costs and the cost of monitoring the agreement.
The companies do not admit wrongdoing or liability, and the plaintiffs do not concede their claims lacked merit, according to the decree. The case would be dismissed, but the parties will ask the court to keep jurisdiction so it can settle disputes. Any disagreement must first go through a meeting between the sides, then court-assigned mediation, and only then to the judge.
The three-year term would be extended by however long the Transcript Enhancement Service is delayed past its development period.
Background: a long road from 2021
DRA and the NAD first sued SiriusXM, Pandora and Stitcher in December 2021. DRA's case page says that at the time the companies did not make transcripts available for any of their podcasts, which meant that more than 48 million Deaf and hard-of-hearing Americans were denied the same access hearing users had.
Radio Ink reported that settlement talks surfaced in 2023 without producing a deal, and the case was reassigned to Judge Vargas in January 2025. The decree says the parties reached the agreement through private mediation with retired Judge Steven M. Gold.
"The NAD congratulates SiriusXM on its commitment to providing high-quality transcripts across its platforms and is thrilled to announce that the 48 million deaf and hard of hearing people nationwide will now have access to podcasts via SiriusXM and Pandora," NAD CEO Nick Kiego said, according to Radio Ink. "We hope that this settlement will serve as a benchmark for other providers of podcast entertainment."
Analysis: why this matters beyond one company
Podcasts are now a major way people get news, politics and culture, and for years many platforms treated transcripts as a nice extra. This decree treats them as part of the product, with standards that can be measured.
The details carry the weight. Automatic transcripts that run every word together, label everyone "Speaker 1," and skip the laughter or the music cue can be technically present and still useless. By requiring named speakers, sound descriptions and formatting, and by tying accuracy to specific tools with yearly checks, the agreement sets a quality floor instead of just a yes-or-no requirement. The "Report a problem with the transcript" link and annual complaint reporting give listeners a way to push back when that floor is not met.
The benefits reach past the Deaf and hard-of-hearing community. Good transcripts help DeafBlind users who read through braille displays, people with auditory processing disabilities, people learning English, and anyone who wants to search or quote what was said. For blind listeners, a clean transcript is also a way to jump to a specific moment with a screen reader instead of scrubbing through audio.
There are limits. The decree's main quality rules apply to podcasts the companies own or license, while third-party shows get Rev-based transcripts and a shared accuracy floor. Human editing covers 30 shows, not the whole catalog. And the agreement lasts three years unless extended, so what happens after that will depend on whether the practices stick.
What's next
The decree takes effect when Judge Vargas signs it. After that, according to the document, the companies have 14 days to give the plaintiffs the style guide for comment, 60 days to share draft accessibility policies, and up to six months to build the Transcript Enhancement Service, with written status updates to the plaintiffs every two months during that period. Once the service is running, the plaintiffs will pick the first 30 shows for human cleanup.
Listeners who find errors will be able to use the "Report a problem with the transcript" link that must appear near transcripts and in the help centers of the SiriusXM and Pandora websites and apps.
Sources
- SiriusXM and Pandora Agree to Provide High-Quality Podcast Transcripts for Deaf and Hard-of-Hearing Consumers — Disability Rights Advocates
- Consent Decree, National Association of the Deaf v. Sirius XM, No. 1:21-cv-10542 (S.D.N.Y.) — Disability Rights Advocates
- National Association of the Deaf v. Sirius XM case page — Disability Rights Advocates
- SiriusXM, Pandora Settle Deaf Podcast Transcript Lawsuit — Radio Ink
- SiriusXM and Pandora will provide podcast transcripts for the Deaf and hard-of-hearing community — Engadget
