Are Authors Lying About Using AI?

Summary of the episode: “Are Authors Lying About Using AI?”

1) Opening discussion: adaptations, translation, and “interpretation” vs. “violation”

The hosts open with a conversation about adaptation and fidelity, sparked by a new discussion around The Odyssey and a London Review of Books critique by translator Emily Wilson (noted for her 2017 translation, described as a leading English standard).

They contrast Wilson’s earlier openness to interpretation (before a film adaptation) with her later article calling the film a violation of what the Odyssey represents. From there, they broaden the theme: people evaluate adaptations differently depending on whether they feel the work is respectful “interpretation” or a betrayal of the original intent.

The episode then pivots from this cultural lens into industry questions—especially whether AI use is being misrepresented, and whether criticism is fair or harmful.

2) Podcast housekeeping and craft/career pressure

Victoria Aveyard and Sman Chenani shift into publishing-life realities: heat waves, touring schedules, edits, and ongoing anxieties about production timelines and promotional workload.

They also introduce weekly “theme” segments—one host’s punishment theme and the other’s grand ambitions—then discuss pressure from:

• deadlines,
• series structure,
• and how publishing economics (profit and loss) shape what gets funded.

A longer segment focuses on managing editing feedback, plotting/structural challenges, and the pressure to “lock in” on craft near publication.

3) Main hot topic: Are authors lying about AI use? (The Atlantic controversy)

The core argument centers on reporting and detection: an Atlantic article that allegedly uses AI-detection research to target a specific self-published book, Daggermouth by H.M. Wolf.

Key claims and concerns

• AI detectors aren’t reliable enough

  • The hosts argue no current tool can definitively prove AI authorship.
  • They criticize overstated detection claims and reject the idea that “AI hunting” is scientifically solid.

• Headline framing despite no hard evidence

  • They note the Atlantic piece allegedly acknowledges no hard evidence, yet the unpaywalled headline/sections still strongly imply AI generation or heavy assistance.

• Non-public data and named targeting

  • The referenced academic paper is said not to name authors/titles publicly.
  • The hosts argue naming likely occurred through raw data shared with the reporter—reducing accountability because readers can’t evaluate the dataset or selection process.

• Questionable ethics/timing

  • They flag odd timing and note that H.M. Wolf is a young Black woman, arguing the targeting and public framing could feel especially harmful and potentially driven by engagement.

• Pattern-level findings vs. individual accusation

  • They acknowledge the research may contain useful signals, such as:
    • growth in self-published output after generative AI became available,
    • potential dilution effects on KU/Kindle Unlimited revenue,
    • category shifts (including romance).
  • However, they criticize converting those broader patterns into a witch-hunt for a single title.

4) What the research does suggest (and what the hosts think it means)

The hosts discuss methods described in the paper, including:

• comparing “rare phrase” overlaps between supposedly AI-written titles and human work,
• and observing that the “best” AI output imitates real writing patterns closely.

Their interpretation is that even if statistical differences exist, that still doesn’t justify confident claims about one specific author/book—because the labeling depends on imperfect detectors.

5) “Hybrid writing” and whether authors can claim authorship

The discussion pivots to a practical/psychological question: even if detection can’t be proven, could many authors be using “hybrid writing”?

• Definition discussed
  • Create an outline,
  • feed it to AI for a draft,
  • then edit extensively to make the result sound like the author—saving time on drafting.

They argue hybrid use may be widespread and can create a belief gap:

• The author may feel they “wrote” and “edited.”
• Critics may see the AI-drafted portion as undermining the honest act of creation.

There’s also debate about whether market pressures push authors toward AI to keep up.

6) Dystopian futures: escalation vs. labeling vs. punishment

They consider two possible futures:

1. AI surpasses human writing

   • If AI-assisted work becomes smoother and dominant, human-only writing may sound worse to readers trained on AI-like styles.

2. Arms race

   • Authors may feel compelled to use AI/hybrid methods to compete, turning writing into a speed/volume contest rather than craft.

Their proposed responses differ

• One host favors a “punishment” model:
  • severe consequences if AI cheating is exposed, to deter others.
• The other host leans toward a categorization/labeling approach:
  • AI-assisted work should be treated as different, not equivalent to fully human authorship, so readers can understand what they’re buying/reading.

7) Generational and emotional framing (AI as “frictionless” creation)

The hosts connect AI adoption to broader “frictionless” trends, including an analogy to Gen Z dating/sex behavior described in a New York Times piece.

Their view: younger creators may see AI as natural because it removes hard steps, but doing so can erase something real about creation—such as emotional labor and meaningful struggle.

8) “Carrot” vs. “stick”: changing the culture without witch hunts

They reject the “Minority Report” style approach of AI-detector witch hunts. They also argue that if detectors improve, AI users will evolve ways to evade them.

One host suggests a “carrot” strategy:

• evangelize the value of making real human art at a human pace,
• and treat creative effort as inherently meaningful—not just output for money or clout.

9) Legal/verification angle: “prove it” through discovery

The hosts argue that truth requires evidence, and that if accusations are made without proof, lawsuits can trigger discovery.

They imply the Atlantic’s claims could be weak enough that the targeted author might sue (or should sue) to force disclosure of methods/data. They also raise a concern that marginalized authors may be less likely—or less able—to sue, making targeted accusations more damaging and asymmetrical.

---

Presenters / contributors

• Victoria Aveyard (presenter/host)
• Sman Chenani (presenter/host)
• Liam Bill (producer/editor)
• Michael Blank (logos/art/graphics)
• Thieves and Louisis Adrien (music)
• June Seekia (production support)