ACX and AI Narration: What the Rule Says and What It Leaves Open

No, not by default. ACX's audio submission requirements say your audiobook "must be narrated by a human unless otherwise authorized", and that unauthorized text-to-speech, AI or automated recordings are prohibited. If you made an audiobook with a third-party AI voice tool, do not upload it to ACX unless Audible has authorized that title or route.

The rule is on ACX's audio submission requirements page, which displays the date April 15, 2026. We read it on November 8, 2026.

What ACX's AI narration rule says

ACX's rule, on the help page displaying April 15, 2026, has four parts that matter to an author:

• The default is human narration. The submitted audiobook "must be narrated by a human unless otherwise authorized".
• Unauthorized synthetic audio is banned. The page says: "Unauthorized use of text-to-speech, AI, or automated recordings in ACX titles is prohibited."
• It reaches the producers you hire. ACX tells rights holders not to pay a producer if they suspect unauthorized text-to-speech.
• Change is signalled but undated. The page says Audible "is working to accept third-party TTS content from publishers and creators who are interested" and will share more on timing and availability.

That last line is why this page has a date on it. As of November 8, 2026, this ACX page says Audible is working on third-party text-to-speech. It does not establish that third-party text-to-speech is generally accepted.

What "otherwise authorized" does not tell you

ACX's page does not define "otherwise authorized" in the text we could read, and it names no approved tools. Treat any list of "ACX-approved AI voices" with suspicion unless it links to Audible saying so.

Two Amazon programs involve synthetic voices, and it is natural to assume they are the exceptions. The ACX rule does not say that, so we don't either:

• Audible audiobooks with virtual voice are created through KDP, not uploaded to ACX. KDP describes them as an invite-only beta for eligible KDP eBooks. Our Virtual Voice page covers eligibility and terms.
• Voice Replica is a beta described on ACX's help site for narrators. ACX asks participants to keep beta details confidential, so we only note that it exists.

The other ACX documents we checked did not fill the gap. The extracted text we reviewed of the Conditions of Use and the content acceptance guidelines contained no statement on AI or synthetic narration. The Conditions refer to Production Standard Terms, which we have not read.

What to do in your situation

Your next step depends on how the audio was made. This table applies the rule as read on November 8, 2026.

If you are unsure whether a specific plan counts as authorized, ask ACX in writing and keep the reply. A vendor's blog post saying its voices are "ACX-ready" is not authorization from Audible. Technical compliance with ACX's audio specs and permission to use the voice are separate questions.

If you already have AI-narrated audio

Mastering does not change how the narration was made, and the rule as worded covers unauthorized text-to-speech and AI recordings. Under the pages read on November 8, 2026, these are the practical steps:

1. Keep it off ACX. Do not ask a producer to deliver it for you either. The rule covers what producers deliver, and ACX tells rights holders not to pay when they suspect unauthorized text-to-speech.
2. Work out what you actually hold. An unmodified LPF package from Google Play Books, ElevenLabs or Spoken Press fits INaudio's digital voice route. Ordinary audio files from a digital voice provider point to Spotify for Authors, whose upload page lists MP3, WAV and FLAC. Those Spotify files will not qualify at INaudio. Selling it yourself is covered in selling audiobooks direct.
3. Do not count on virtual voice to rescue it. KDP's virtual voice creates a new audiobook from an eligible eBook. Nothing on the pages we read describes bringing in audio you already made.
4. Check exclusivity before you place it anywhere. ACX's distribution page, displaying October 21, 2026, says exclusive titles may not be distributed or sold on any site or store outside Audible. It does not address a second recording of the same book made separately, so ask ACX before running a human edition there and an AI edition elsewhere. If the eBook is in KDP Select with a virtual voice edition, KDP says the exclusivity covers any audio edition.
5. Keep your project files and receipts. ACX says it will share more about third-party text-to-speech. If a route opens, you will want the original exports and the tool's licence terms, whatever the new conditions turn out to be.

Recording it yourself or hiring a narrator through ACX

Human narration is the route ACX's rule plainly leaves open, whether you read the book yourself or hire someone. ACX says it is open to residents of the United States, United Kingdom, Canada and Ireland, on a page displaying January 7, 2026.

Recording it yourself

If you narrate, the requirement to be human is met, and the audio specification becomes the hurdle. ACX's April 15, 2026 page sets limits on format, loudness, peaks, noise floor and room tone, which our audio specs page lists in full. Build your first chapter to that spec before you record the rest.

For time, ACX's blog post of June 11, 2013 gives about 9,300 words per finished hour and about four hours of work per finished hour. Estimate only, for an 80,000-word book: 80,000 ÷ 9,300 = about 8.6 finished hours, and 8.6 × 4 = about 34 hours of work. Those are rules of thumb from 2013; your own time may differ.

Hiring through ACX

ACX's royalties page, displaying July 31, 2026, sets out three ways to pay a producer:

Rights holders earn 50% exclusive or 30% non-exclusive under the new model, which applies from May 26, 2026 to newly claimed titles, new users and enrolled titles. The legacy model pays 40% or 25%, and ACX says it ends at year-end 2026. These are contractual rates on sales under ACX's own formulas, not a share of your list price, so model them with the audiobook royalty calculator instead of multiplying by the price.

A Pay-for-Production budget has to come from quotes. The only ACX figure we have is that 2013 post, which estimates about $200 per finished hour for narration plus about $200 for post-production. Estimate only, on 2013 figures: 8.6 × $400 = about $3,440. Treat it as a historical reference point, and see human versus AI narration cost for the wider decision.

Exclusivity changes what else you can do. ACX's distribution page says non-exclusive titles may be sold elsewhere, that ACX does not distribute to libraries, and that Audible holds distribution rights for seven years under either option.

Human-narrated audio can also reach Audible outside ACX. INaudio says it can deliver to Audible if an eBook is on Amazon and the title is not claimed on ACX, and allows about 45 to 60 days for processing.

What to put in writing with a narrator

ACX's warning about unauthorized text-to-speech lands on you as the rights holder, so settle this before work starts. These are our suggestions. ACX does not prescribe this wording:

• All narration, including pickups and corrections, will be performed live by the named human narrator.
• No text-to-speech, AI voice or voice replica will be used in any delivered file without your written agreement and Audible's authorization.
• The narrator will tell you before using any AI tool on the recording.
• You may ask for raw session files if you have a concern about a chapter.

Questions the rule does not settle

ACX's page is short, and several questions authors ask have no answer in the text we could read on November 8, 2026. Silence here is neither permission nor prohibition.

• A book that is partly AI. The text has no exception for a mostly human recording with AI pickups or corrected lines. The safe assumption is that the rule still applies, and ACX can confirm.
• AI tools used only for clean-up. The rule names "text-to-speech, AI, or automated recordings". It does not separate AI noise reduction or mastering from AI narration. Ask ACX about the specific tool if you rely on one.
• A narrator offering a replica of their own voice. We found no exemption for this in the rule. Treat it as covered unless ACX confirms otherwise in writing.
• How ACX detects AI, and what happens if it does. The text describes no detection method and no penalty. Do not read that as safety.

Where AI-narrated audiobooks can go instead

Other stores have published their own rules on digital narration, and they differ. As read on November 8, 2026:

• Spotify says it accepts digital voice narration and adds a notice to the description.
• Voices by INaudio accepts digital voice narration only in original, unmodified LPF packages from three named providers.
• Kobo asks you to label the narrator as a synthesized voice, according to its audiobook FAQ as captured on November 8, 2026. A follow-up read that day was blocked, so check the live page.

The full comparison of which retailers accept AI narration shows each rule beside its source and date, including the stores where we could find no rule. For money and reach, see audiobook distributors compared.

What we could not confirm

We can only report what ACX's published pages show, and five things stayed out of reach on November 8, 2026:

• What "otherwise authorized" covers. ACX's page does not list it in the text we could read.
• ACX's Production Standard Terms, which we did not read.
• When Audible will open third-party text-to-speech, or on what terms. ACX gives no date.
• How ACX enforces the rule, including any detection method or penalty.
• Whether an exclusive ACX title permits a separately recorded AI edition elsewhere.

Our reading of ACX's help pages came through a page-extraction tool rather than a saved copy of the page, so check the live page before you rely on the exact wording. We log rule changes in the AI narration policy changelog, and every retailer rule we track starts from the AI narration rules hub.

This is general information, not legal advice.

Last verified 2026-11-08. Rules and prices change: check the linked primary page before you rely on this.