1. Purpose
To provide a controlled framework for the responsible use of artificial intelligence in publishing, ensuring that AI is used as an assistive technology rather than a substitute for human authorship, while maintaining transparency, accountability, factual accuracy, intellectual-property protection and confidentiality.
2. Acceptable use
Artificial intelligence may be used within research, writing, editing, translation, design and other publishing processes.
Tuwhiri accepts the responsible use of AI in accordance with the content guidelines of IngramSpark and Lightning Source, through which Tuwhiri books are printed and distributed.
A work may be considered for publication where:
- generative AI has been used only in an assistive capacity;
- most of the work has been written or created through human creativity and contribution; or
- AI-generated material, including translation or images, has received significant and material human editorial review and contribution, with proper attribution.
Assistive use may include research support, planning, editorial assistance, proofreading and revision.
AI-assisted or AI-generated wording may remain in a manuscript where it has been sufficiently reviewed and developed by the author or editor to meet these requirements.
Where a work sits between these categories, Tuwhiri’s editors will judge whether the human contribution is substantive enough that the named author can honestly claim authorship of the finished work.
These categories follow Ingram’s current Generative AI Guidelines.¹
3. Author and editorial responsibility
The named author and Tuwhiri’s editors remain responsible for the finished publication.
Authors must inform Tuwhiri, at submission, whether and how generative AI was used in preparing the work. Where the use goes beyond assistive support, the finished work must acknowledge it – in the preface, acknowledgements or colophon, as appropriate.
Information obtained through AI must be checked before publication. Quotations, references, factual claims and biographical information must not be included unless they can be verified.
The finished work must comply with copyright, trademark, privacy, publicity and other applicable rights. The use of AI does not remove the need to obtain permission for protected quotations, images, adaptations or other material.
Tuwhiri’s editors will not submit an unpublished manuscript to a third-party AI tool without the author’s consent, and then only via services that do not retain the material or use it for training.
4. Content that will not be accepted
Tuwhiri will not publish:
- works created solely through generative AI;
- works that infringe intellectual-property or copyright law;
- automatically or repetitively mass-produced content;
- unauthorised summaries, workbooks or adaptations of another author’s work;
- content that misleadingly imitates another title, cover, author or creator; or
- content or metadata that is misleading or likely to confuse readers.
These restrictions reflect Ingram’s Generative AI and wider Content Guidelines.
5. Attribution and metadata
Authorship and contributor information must accurately represent the creation of the finished work. Generative AI will not be credited as an author.
Where generative AI has contributed to a work in a way that requires attribution, Tuwhiri will provide appropriate attribution in accordance with Ingram’s current guidance.
For translated works, IngramSpark currently advises publishers to include the translator as a contributor. Where AI has been used for the translation, ‘Artificial Intelligence’ should be entered as the translator’s name in Ingram metadata, as Ingram’s guidance specifies.
Elsewhere – in the work’s own acknowledgement of AI use (see section 3), in Tuwhiri’s records, and in any channel that does not mandate particular wording – AI contributions will be attributed with the briefer and more specific tag ‘AI:{model}’ (for example, ‘AI:gpt-5.6’ or ‘AI:fable5’), listing each model where more than one was used.
Where Tuwhiri distributes a title through channels other than IngramSpark – for example, Amazon KDP, which requires AI-generated content to be declared at upload – those channels’ AI-disclosure requirements will also be met.
6. Editorial process
AI-assisted manuscripts will proceed through Tuwhiri’s usual editorial process. The use of AI is not in itself a reason to reject a manuscript.
The relevant consideration is whether the finished work falls within Ingram’s acceptable AI-content categories and complies with its wider requirements concerning originality, accurate presentation and intellectual-property rights.
7. Review
Tuwhiri will review this policy if IngramSpark or Lightning Source materially changes its content or generative-AI guidelines.
IngramSpark retains discretion over whether content is suitable for its catalogue and notes that a title may not be accepted by every distribution channel.
IngramSpark’s enforcement is partly automated, and titles have been flagged in error – including at least one book removed for merely discussing AI, later reinstated on appeal. If Ingram flags or removes a Tuwhiri title that complies with this policy, Tuwhiri will appeal the decision and support the author through the process.
8. Out of scope
The use of Tuwhiri publications as training data for AI systems – including text and data mining licences and any opt-in or opt-out arrangements offered by distributors – is a separate matter, to be addressed in author agreements rather than in this policy.
This policy will be reviewed in August 2027, or earlier if IngramSpark changes its guidelines.
Image credit: AI:grok-imagine-image
Note
1. IngramSpark publishes its Generative AI guidelines (acceptable and unacceptable AI content) on the Lightning Source content-guidelines page:
https://www.ingramcontent.com/page/content-integrity-guidelines
The IngramSpark-branded Catalog Integrity Guidelines page from October 2023 at is:
https://www.ingramspark.com/blog/ingramsparks-catalog-integrity-announcement
It reproduces only the blanket prohibition on ‘content created using automated means’, without the acceptable-use carve-outs. Both pages are IngramSpark policy; the fuller Lightning Source version is the operative one for this policy.