AI Actions: Usage calculation

The AI Actions feature enables Customers to automate content management tasks using AI, allowing them to choose their preferred models and configure custom prompts.

AI Actions can be utilized across various areas of Contentful’s product offerings, including transforming fields and entries (e.g., translation, rewriting) or modifying other entities such as content types, taxonomies and more. Additionally, AI Actions can operate across multiple entities.

Customer will have a specific AI Consumption Units (described below) limit per year as set forth in the applicable Service Order or the Feature Descriptions linked therein. “Words Generated” are counted each time Contentful’s system is triggered to interact with the AI model, these models receive “prompts” and then return output as “responses”, and the number of Words Generated is calculated from the response text.

Each Word Generated (and any other outputs made available by Contentful in the future) will consume Customer’s AI Consumption Units capacity in accordance with the table below. These generations could be triggered when a user manually triggers an AI Action or when it is triggered by a workflow or automation capability, which can be configured by the user. All generated content will count towards capacity; this includes manually triggered computations that are not accepted, saved or published as well as any content generated in a playground or sandbox environment.

The calculation of the number of Words Generated in a given output text from AI Actions is based on the methodology implemented by the applicable word-counting library (“Library”). The Library determines word boundaries using a bitmap-based approach, which efficiently identifies separators such as spaces, punctuation, and specific Unicode characters across multiple languages. The total number of Words Generated in a given output text from AI Actions is determined by counting the segments of text that are separated by recognized word boundary characters.

The number of Words Generated may vary depending on language-specific rules, character structures, and formatting. For example, in English and most Latin-based languages, words are typically separated by spaces or punctuation, whereas in languages such as Chinese, Japanese, and Thai, words are often inferred based on character groupings rather than explicit spaces. The Library processes text accordingly to support accurate segmentation.

Additionally, numbers, punctuation, and special characters may influence the total Words Generated count. In certain cases, sequences of numbers formatted with commas, periods, or spaces may be counted as multiple Words Generated, while contiguous alphanumeric sequences without spaces may be counted as a single Word Generated.

For example:

  • “1,234” is counted as 2 Words Generated.

  • “020 1234 1234” is counted as 3 Words Generated.

  • “+44 (0) 20 1234 1234” is counted as 5 Words Generated.

For non-Latin scripts, Words Generated may differ depending on word segmentation conventions.

For example:

  • The phrase “2025在一个宁静的小村庄里” is counted as 10 Words Generated.

  • The phrase “在一个宁静的小村庄里2025” is counted as 11 Words Generated due to the placement of the number.

The total Words Generated count includes all visible words within a given text, as well as any words that may be present in structured data formats such as JSON, Markdown, or RichText. In the case of structured text, formatting elements (e.g., markdown syntax or JSON keys) may contribute to the overall Words Generated count.

For example:

  • A plaintext version of a recipe may contain 55 Words Generated, whereas the same text in a RichText format may contain 431 Words Generated due to additional metadata.

If word counting is performed as part of a multi-step process where intermediary text is generated to produce a final result, all Words Generated during intermediary steps are included in the total Words Generated count, even if they are not directly visible to the end user. For example, if an initial step generates bullet points and a subsequent step transforms them into a paragraph, both the bullet points and the paragraph contribute to the total Words Generated count.

If Customer exceeds the applicable capacity threshold, notification may be provided, and further usage may be subject to additional fees.

Any modifications to the Words Generated calculation methodology or processing rules will be communicated as necessary.