- Generative AI features in Acrobat
- Adobe Generative AI user guidelines
- Content usage and handling practices
- Before you begin
- Get AI-generated answers
- Get AI-generated overviews and summaries
- Use the generative AI features from Acrobat Home
- React to the content responses
- Turn off the generative AI features
- Generative AI features in Acrobat mobile
- Known issues
- Usage limitations
AI Assistant and generative summary in Acrobat have certain usage limitations. As an Individual plan user, users are provided with a limited number of Requests for AI Assistant in Acrobat or Reader at no additional charge, after which users must purchase an AI Assistant for Acrobat subscription to continue using these features. These Requests cannot be used towards or in combination with Creative Cloud generative credits. We reserve the right to update usage limits or to throttle or suspend usage in excess of our limits at any time.
AI Assistant capabilities are available to customers with a free or paid entitlement when they purchase a new AI Assistant for Acrobat add-on subscription. AI Assistant for Acrobat subscriptions are available on desktop and web in English—other languages to follow. AI Assistant on Reader mobile (beta) is available in English at no additional charge for a limited time. Paid subscriptions for Teams and Enterprise customers will be coming soon.
Tiers
Limited Use: users may be allotted a limited number of Requests for AI Assistant and generative summary to try out the service.
- Acrobat for Individuals: all signed-in users of free and paid plans will receive 5 lifetime Requests for Limited Use.
Full Access: users may choose to purchase the AI Assistant for Acrobat plan to receive more Requests to use with AI Assistant and generative summary. See plans information here.
- Acrobat for Individuals: users that purchase AI Assistant for Acrobat plan will receive 1,000 Requests/month per user. After consumption of 1,000 Requests/month per user, Adobe may throttle usage.
Beta, Prerelease, and Early Access Programs: you will receive a complimentary number of Requests for AI Assistant and generative summary in exchange for providing feedback about the service. AI Assistant in Acrobat mobile (iOS) is currently made available as part of a consumer-facing early access program, and its Android equivalent is in beta. Teams and Enterprise customers can also try out Generative AI features by requesting to join the Technical Evaluation Program or Early Adopters Program. Early access programs are not a part of the AI Assistant for Acrobat paid plan.
Definitions and limitations
Request: this occurs when a user submits a textual prompt, selects a suggested prompt, or initiates an AI-related task that conforms with prerequisites described here. A Request must result in a Response to be counted towards usage limits of AI Assistant for Acrobat.
- For example, a Request may be when a user types a question, submits it to AI Assistant, and receives a Response.
- As another example, a Request may be when user selects the Generative summary action and a summary is generated.
Response: the successful generation of a single summary, task, or answer to a Request. A Response does not include when the Request fails because of technical limitations described here.
Common questions
What is throttling?
All Requests made to AI Assistant for Acrobat are monitored to protect system resources and preserve our ability to serve as many users as possible.
The rate of resource consumption by the same consumer (for example, the same user ID, IP address, agreement IDs, etc.) is limited by the minute, hour, and day. When a consumer crosses a throttling threshold, that request is rejected when there are too many Requests submitted (response status code HTTP 429).
What is the allowed rate of requests?
Requests higher than 1,000 per month, per user, may encounter throttling policies. Throttling policies are based on historical daily request data, including legitimate usages that stress the system.
Each request to AI Assistant for Acrobat is evaluated based on the amount of system resources it consumes. Different parameters passed to the same endpoint might contribute to a different amount of resource consumption.