Here’s my disclosure: I didn’t write this blog post with the help of AI.
For byline articles and other contributed content, this type of disclosure — whether AI was used in part or in whole — is becoming more common. It is a concern we are increasingly flagging for clients, particularly those in the legal space as related bar rules over AI use continue to evolve.
No author wants to sink significant time and energy into an article only to have it rejected for running afoul of a publication’s AI-content policies. Here are some helpful insights for hopeful authors to keep in mind.
AI Disclosure Is Becoming the New Norm
Here are some examples of publications that already have notable disclosure clauses:
Bloomberg Law
You must disclose whether you’ve used AI in any part of the drafting or writing process.
Law360
The article must be written by the individual(s) named on the byline. Artificial intelligence tools may be used for research and refining, but the author must independently verify the accuracy of all content. Additional use of AI must be disclosed to Law360.
HR Daily Advisor
AI-generated or AI-assisted content must be clearly disclosed and human-reviewed.
HR Dive
The use of generative artificial intelligence is not permitted in the writing of op-eds. Any submissions that are AI-generated will be rejected.
Entrepreneur
We do not accept articles that feature AI-generated text. Articles are run through an AI detector to identify such articles and will be flagged accordingly.
Fast Company
We do not publish AI-generated articles. All editorial content published by Fast Company must be reported, analyzed, and written by a human author. Generative AI tools may assist with brainstorming, research support, organization, copyediting, or other productivity tasks, but AI may not serve as the primary creator of any content.
Lab Manager
Any use of AI must not breach journalistic ethics, and plagiarism will not be tolerated — Lab Manager’s editorial team runs all contributed content through trusted plagiarism detection software as part of the publication’s standard review process. Content submitted to Lab Manager must be the author’s own, and not present others’ data, words, or imagery without seeking permission (if applicable) and providing appropriate citation.
What does Disclosure Require?
Accurately stating that content is AI written is a tricky issue. It is a stinging accusation to suggest that a piece of content is largely the work of AI, which is why we lean on upfront guidance to authors before drafting occurs.
We’re nearing the point where a submitted draft is first reviewed by an AI detector and subsequently rejected by a human editor based on the belief that it is more than 50% synthetic, in violation of the outlet’s author agreement. We then have a duty to advise the marketing/communications team and/or the author that their work will not be accepted, even after substantial time and effort has been invested in creating the piece. No one wants that outcome.
If content is wrongly flagged, a prior history of honesty and disclosure with an outlet will go a long way in overcoming skepticism. Conversely, asserting that a piece is human and having it rejected for lack of disclosure can dim future publishing prospects — for the author, the submitting organization and, frankly, for us as outside consultants.
Professional services organizations should be addressing this issue as part of their overall policies regarding AI use and scrutiny of unmonitored “Shadow AI” use.
Should You Use AI in Drafting an Article?
We’re not the lawyers, we just work with a lot of them. Here are some key considerations:
Clear-Cut AI Disclosure Moments:
- The article was drafted conversationally with an AI bot.
- Someone on the team used AI in the drafting or editing process.
Gray Areas for AI Disclosure:
- AI is used to research a topic, but not in the subsequent drafting of the content.
- Prior, personally generated content is used as a “walled garden” of source material to construct a new draft.
- AI was used to proofread an article that is 100% human-written, and some of the edits are incorporated into the final version.
Existential AI Disclosure Questions:
- If AI did the first draft or reviewed and edited the first draft. Is this now sufficiently “human?”
- If AI drafts all or part of the article, should the tool be listed as co-author?
How and How Well is AI-Generated Content Detected?
Since the “a-ha” moment in 2022 when ChatGPT debuted, there have been ample sites offering to “detect AI.” In general, trust in their results has been fairly low.
People have also deduced — rightly or wrongly — that content is AI based on tells like em dashes and repetitive, overly flowery content.
Sometimes the tell is obvious, like links with “ChatGPT” tracking code visible or moments like this one where a legislator read part of the bot’s prompt into the record.
Recently, there has been a lot of buzz around Pangram, which is leading the market in terms of AI detectors and is considered generally accurate, although not perfect.
We tested it ourselves by creating and running AI content through Pangram. The results were promising with a spot-on deduction of where and when AI was used. We can assume most, if not all, leading news outlets will be incorporating similar rigorous detectors to match their disclosure policies.
With the LLM models advancing by leaps and bounds nearly every week, less emphasis will be on “too many em dashes” and more will be on honest disclosure and accurate detection of bot content.
AI Disclosure Takeaways
- Understand AI disclosure is increasingly a part of author agreements. Be sure you look at the most recent version and carefully review AI provisions before you begin drafting.
- AI detectors are getting much better. Consider experimenting with one to get a feel for the sensitivity.
- Consider and, where unclear, ask the outlet what level of AI use requires disclosure.
- Have strong internal policies that regulate AI use and disclosure in content creation.
- Err on the side of transparency. Build and maintain trust with internal/external comms teams and outside media outlets.
This is a shifting landscape we are closely monitoring.