Series Bible Continuity Errors: What AI Caught in My Own Manuscript

The continuity error you are carrying in your series right now is probably not the one you are watching for.

It is not the eye colour you changed between books. You have a spreadsheet for that. It is not the character's age on their birthday in chapter fourteen. You checked that twice.

It is the assumption so embedded in your understanding of the story that you have stopped being able to see it at all. Your brain fills in what it knows should be there, reads past the gap, and moves on. Fourteen re-reads later, it is still there.

I built Editorial Conductor partly to test whether a system with no prior knowledge of your story could catch that kind of error. A system with no cached version of what the chapter was supposed to say. So I ran an experiment.

The experiment

I have been working on a literary fiction series called ORAI. The first book is set in Porto. The protagonist is Tomás Ferreira: 38 years old, a book restorer, close third person, past simple tense throughout. The series bible is explicit about what cannot change.

I wrote a second version of the opening chapter that broke every rule in that bible. Not subtly. Aggressively.

The new chapter introduced a character named Marco. Twenty-five years old. Running a productivity app. Narrated in first person, present tense. Set in a WeWork in Lisbon. Startup culture. Exclamation marks. Social media language. Every drift signal the bible explicitly forbids.

Then I ran both chapters through the same ten-agent pipeline, with the series bible attached to both.

The conforming chapter scored 7.8 out of 10. The drift chapter scored 2.6 out of 10.

Here is what the agents caught in the drift chapter, and why those catches matter.

Finding 1: Wrong protagonist identity

The Literary Agent Simulation was the most direct about it:

"This chapter belongs to an entirely different book."

Six agents flagged the same root problem independently: Series Continuity Keeper, Structural Architect, Voice Consistency, Literary Agent, Award Jury Reader, and The Conductor. The protagonist in the chapter was simply not the protagonist established in the series bible.

In a constructed test, this looks obvious. It is less obvious when the problem is not a complete protagonist swap but a gradual drift over the course of a book. A character whose decisions stop being consistent with who they were in chapter one. A voice that quietly shifts register by chapter nine. Someone who begins reacting to things in ways that no longer match their psychology.

Agents cannot catch that kind of drift from a single chapter. But they establish a baseline, and the baseline is what makes a second pass useful.

Finding 2: Wrong city, seven times

The bible states Porto only. Never Lisbon.

The drift chapter opened in a WeWork on Avenida da Liberdade. Every subsequent reference, seven in total, was flagged on first occurrence and each recurrence.

This matters because real continuity errors rarely arrive as single clean violations. They propagate. A wrong city in one chapter becomes a wrong travel route in the next, a wrong landmark in the chapter after that. Catching the first instance is the one that costs the least to fix.

Finding 3: POV and tense

The bible requires close third person, past simple, throughout Books 1 to 3. First-person narration is explicitly forbidden.

The drift chapter opened: "My name is Marco. I'm 25. I run a productivity app."

This was flagged by the Voice Consistency Agent before the Continuity Keeper even reached it. Voice drift often sits upstream of continuity in the detection order. The agents hear the register change before they can articulate which rule it violates.

The fix note from the Continuity Keeper was precise: rewrite from scratch with Tomás as the subject, remove all first-person references, all Lisbon references, all startup-world references. Not a patch job. A root-cause failure.

That last part is the one worth remembering. Not every continuity error can be fixed by changing a name or a city. Some require acknowledging that the draft went in the wrong direction before the first sentence.

Finding 4: Tone drift signals

The bible has a section most writers skip when building series bibles: explicit drift signals. Things the prose must never do. For ORAI the list includes slang, exclamation marks, startup culture references, casual register, and social media language.

The drift chapter triggered all five. The Award Jury Reader's note was the one that stung:

"Casual register, exclamation-heavy lines, startup jargon, and social-media register where the bible forbids them."

This is not a continuity error in the traditional sense. No character name is wrong, no date contradicts an earlier chapter. But it is the kind of drift that erodes a series quietly over time. A reader cannot name it. They just feel that something changed and they cannot tell you which chapter it started in.

Finding 5: The series object is missing

The Award Jury Reader flagged something none of the other agents mentioned. The 1887 book with the word in the margin, the central object of the ORAI series, does not appear in the drift chapter at all.

"No mention of the central object that should drive the series narrative. The inciting object has vanished completely."

The drift chapter had replaced the manuscript's actual subject with a productivity app. The series object had not been mentioned because there was no place for it in a narrative about someone else's startup.

This is the hardest category of continuity error to catch yourself. Not something that changed, but something that was supposed to be there and is not. Absence is invisible on a re-read. It only surfaces when you are comparing the chapter against a document that specifies what must be present.

Finding 6: A name that should not appear

This one came from the conforming chapter, not the drift chapter. One line: He had not been in a rush since Clara left.

The bible states that Clara must not be named in Books 1 or 2. An oblique reference is permitted, "since she left," but not the name itself.

The chapter scored 7.8. The Clara flag was one of two Critical findings in an otherwise clean draft. Six agents saw nothing structurally wrong with the prose. One named reference, planted without thinking, would have created a continuity problem I would have had to unwind across two books.

Why re-reading does not catch this

The experiment was designed to be extreme. A real manuscript does not drift this completely from its own bible. But the mechanism that caused me to miss the Clara reference in the conforming chapter is the same mechanism that lets subtler errors accumulate across a series.

Your brain is not a continuity checker. It is a story-completion engine. It reads for meaning, not for compliance. It fills in the rules it knows you intended and moves on.

A system that has never read your manuscript before, that has only the chapter and the bible, has no assumptions to fill in. It compares what is in the text against what the bible says should be there. That is not better than a human editor's judgment. It is a different kind of attention. The kind that does not get tired, does not remember what you meant to write, and does not skip past a name it has seen ten thousand times.

---

The full ORAI experiment is live on the demo page if you want to see the complete agent-by-agent scores and the Conductor brief before uploading anything of your own.

If you have a bible and want to run your first chapter against it, five free credits are waiting. No card required.