The short answer
Paste one point-of-view label per chapter in reading order, choose the rule your book follows, and the checker maps every chapter to its head, counts each POV share, counts switches between chapters, and flags single-visit POVs plus mid-list jumps. Under one POV per chapter, calm rotations with stays of two or three chapters pass, restless switching earns a warning, and a list crowded with heads that appear once fails. Scene-break switching gets more room because a visible break prepares the reader, and omniscient narration passes freely since its narrator is allowed inside every head. The scores on this page are deterministic heuristic checks against stated craft conventions, not grades for your writing, and the tool never reads your prose.
Why slips hide in chapter lists
A single chapter almost always feels fine on its own. The point of view inside chapter nine is clear while you are drafting chapter nine, because you know whose eyes you chose that morning. The problem only appears across the sequence: chapter nine belongs to the detective, chapter ten opens inside the suspect for two pages, chapter eleven returns to the detective, and no single chapter looks broken. Readers experience the book as a sequence, so the unit that matters is the list, not the chapter.
Chapter lists hide three specific faults. The first is the quiet cameo, where a minor head carries one chapter in the middle of the book and never returns. The second is the restless rotation, where every chapter changes heads and the reader never settles anywhere long enough to attach. The third is the drift, where the opening promises one head and the middle slowly hands chapters to newcomers until the original lead holds a minority of the book. None of these show up in a spellcheck or a word count. They show up when you line up one label per chapter and look at the pattern the way a reader lives it.
This checker exists for that lining-up step. It cannot tell you whether a chapter is well written, and it does not claim to. It tells you whether the assignment of heads to chapters keeps the promise your narration made in chapter one.
Labeling your chapters before you paste
Use one short label per chapter, in reading order, separated by commas. Short names work best: Maya, Jon, Priya. Single letters work while you are testing, but real names prevent mistakes once the draft is long, because a letter you reuse for two different people will corrupt the count silently.
Repeat the label exactly whenever a head returns.
The three rules and what each promises
One POV per chapter is the strictest common promise in modern commercial fiction. Each chapter stays inside a single head from its first line to its last, and the chapter break is the only place the head may change. Most romance, mystery, thriller, young adult, and single-protagonist fantasy runs on this rule. Pick it whenever your chapters each name or imply one viewpoint character and never leave that head mid-chapter.
One POV per scene allows switches inside a chapter provided a visible scene break separates them. Chapters still open in one head, but a centered break or a new section can hand the scene to another head, as long as no single scene mixes heads. Pick this rule when your chapters are long and multi-scene, common in epic fantasy, historical fiction, and some science fiction. Note what the checker can and cannot see here: it reads chapter-level labels only, so it treats a chapter as one assignment and judges the switching pace leniently, on the assumption that scene breaks inside chapters carry part of the load.
Omniscient narration makes no limited promise at all. The narrator stands outside the cast, reports thoughts from any head, comments, summarizes, and may move between minds within a paragraph. Pick this rule only if your narrator does this openly and consistently, with one steady guide voice across the whole book. Do not pick omniscient as a way to excuse three slipped chapters in an otherwise limited book; readers feel the difference between a designed guide voice and a leak.
How the tool math works, step by step
The checker runs five calculations on your pasted list, always in the same order, with fixed cutoffs. Nothing is random, nothing adapts to genre, and the same list always earns the same verdict.
First it parses. It splits your paste at commas, trims spaces, drops empty entries, and keeps the first sixty labels. A paste of A,A,B,A,C,B,A,A,B,C,A,B becomes twelve chapters: six for A, four for B, two for C.
Second it counts shares. Each POV count is divided by the chapter total and rounded to a whole percent. In the twelve-chapter default, A holds six of twelve for 50 percent, B holds four of twelve for 33 percent, and C holds two of twelve for 17 percent. Shares answer the attachment question: who does the reader live with most.
Third it counts switches. It walks the list and counts every place where a chapter head differs from the previous chapter head. The default list switches nine times across eleven gaps: chapter two to three, three to four, four to five, five to six, six to seven, eight to nine, nine to ten, ten to eleven, and eleven to twelve stay or change as labeled, giving nine changes out of eleven, which is 82 percent. The switch rate answers the restlessness question: how often the reader is asked to re-anchor.
Fourth it finds single-appearance POVs, the one-scene wonders. Any label with a count of exactly one is listed by name. The default list has none, since C appears twice. A list like A,B,C,D,E,A would flag B, C, D, and E as singletons.
Fifth it converts the pattern into risk points. A switch rate above 70 percent adds two points, above 50 percent adds one. Two or more singletons add two points, exactly one adds one. Five or more distinct POVs add one more. Under one POV per chapter, four or more points fails, one to three warns, and zero passes. Under one per scene the same points only warn at four or above, because scene breaks absorb some switching. Under omniscient the points are reported but the verdict stays passing, because switching is legal there.
Apply that to the default twelve-chapter list under one per chapter. Switch rate 82 percent is above 70, so two points. No singletons, so zero. Three distinct POVs, so zero. Total two points, which lands in the warning band. The verdict headline reports twelve chapters with nine switches, the body names the shares, and the recommendations target the switching pace rather than any single head.
Now change one fact and watch the arithmetic move. Take A,A,A,B,B,B,A,A under the same rule. Eight chapters, two POVs, shares 62 and 38 percent. Switches happen only at chapter three to four and chapter six to seven: two switches across seven gaps, 29 percent. No singletons, two distinct heads. Zero risk points, clean pass. The repair lesson is visible in the numbers: bundling stays of three turned an 82 percent switch rate into 29 percent without removing any POV.
Take the opposite extreme: A,B,C,D,E,F under one per chapter. Six chapters, six POVs, every share 17 percent. Five switches across five gaps, 100 percent. Four singletons beyond the returning A, actually five single-use heads besides none returning twice except A appears twice, so B, C, D, E, F appear once each: five singletons. Two points for the rate, two for the singletons, one for six distinct heads: five points, fail. The numbers describe exactly what a reader feels: nobody to attach to, constant re-anchoring, five heads that never earn their visit.
Reading the matrix table
The table lists every chapter in order with three columns: chapter number, POV label, and what changed since the previous chapter. The first row always reads Opening POV. Later rows read Same as chapter plus the number, or Switch from plus the previous head name. Scan the third column first. Long runs of Same rows are stays, and they are good. Clusters of Switch rows are restless passages, and they deserve a second look in the draft.
If switches bunch in chapters six through nine, that passage is the repair site, and chapters one through five can stay untouched.
Illustrative hypothetical: the mystery with a suspect chapter
Consider an illustrative hypothetical novel, a twelve-chapter mystery told by Detective Ruiz. Eleven chapters carry the label Ruiz, and chapter seven carries the label Vane, the suspect, covering the night of the crime from inside his head. The shares read 92 and 8 percent. There are two switches across eleven gaps, 18 percent. One singleton: Vane. Total risk is one point, a warning under one per chapter.
The warning is correct and useful, because the question is real: does the book need seven inside the suspect. Three honest repairs exist. Give Vane a second chapter late in the book, perhaps the interrogation from his side, so the singleton flag clears. Or move the night-of-crime material into Ruiz by letting her discover it: a witness account, a recovered recording, a walk-through of the room. Or cut the chapter and let the gap itself create suspense.
Illustrative hypothetical: the romance with an epilogue head
Consider a second illustrative hypothetical, a ten-chapter romance alternating strictly between Noor and Ellis: Noor, Ellis, Noor, Ellis, and so on through chapter nine, with chapter ten, the epilogue, labeled Sam, the couple child speaking years later. Shares are Noor five, Ellis four, Sam one. Switches run at nine across nine gaps, 100 percent, because strict alternation changes heads every chapter. One singleton: Sam. Two points for the rate plus one for the singleton makes three: warning.
Two faults hide inside one verdict here, and they need different tools. The alternation rate is structural: every-chapter rotation between two heads always scores 100 percent switching, yet many published romances do exactly this and readers love it. The checker flags it because the pace is objectively fast, but the repair is judgment, not arithmetic: if each Noor and Ellis chapter ends on a small pull and opens with a clear re-anchor, keep the rhythm and accept the warning as information. The epilogue head is the sharper problem. Sam appears once, in the final pages, asking the reader to bond with a new voice exactly when the book should be paying off two old ones. Reassign the epilogue to Noor or Ellis looking back, or age the epilogue into an unvoiced time-skip inside an established head, and the singleton clears while the beloved alternation stays. That is the general principle: negotiate with switch-rate warnings, but treat singleton warnings as near-orders.
Repairing one-scene wonders
A head that appears once fails the reciprocity the book sets up. Every chapter spent inside a head asks the reader to learn a voice, a wound, and a want, and a single chapter cannot repay that investment. Under limited narration, cut or expand every singleton; leaving one in place should be a conscious exception with a written reason.
Expansion means a genuine second chapter, not a longer first one. The returning visit must contain information unavailable anywhere else: a decision only this head can make, a room only this head can enter, a secret only this head knows. Place the two visits far enough apart that the reader feels return rather than adjacency, ideally with at least two chapters between them.
Reassignment means finding the singleton chapter material a home inside an established head. List the chapter facts that must survive: plot events, reveals, emotional beats. Then ask which returning head could plausibly witness, discover, or be told each fact. Witness scenes become new scenes for the host; discoveries become investigation; told facts become dialogue with resistance, since people rarely hand over secrets cleanly.
Cutting means deleting the chapter and checking what breaks. Often surprisingly little breaks, because singleton chapters tend to repeat information the main heads already carry. When something does break, usually one reveal, relocate that reveal into a host chapter climax where it hits harder for arriving late. Record the cut in your revision log with the chapter number and the new home of each salvaged fact.
Repairing fast switching without losing energy
Fast switching usually comes from short stays, not from too many heads. Three heads rotating every chapter feel frantic; the same three heads in stays of three feel stately. The cheapest repair is bundling: take chapters four, five, and six, currently split across three heads, and ask whether the middle chapter middle events could be witnessed by the head of chapter four or delayed into the head of chapter six. Each successful bundle removes two switches at once.
Protect the smallest head while bundling. Authors repairing pace often starve the third POV down to one chapter and accidentally manufacture a singleton. Set a floor: every surviving head holds at least two chapters, and the smallest share clears roughly 15 percent of the book. In a twelve-chapter book that means no head below two chapters. If a head cannot hold two chapters of unique material, it is a reassignment candidate, not a bundling victim.
Re-anchor every remaining switch deliberately. Open each new-head chapter with the head name or an unmistakable marker, and close the previous chapter with a small pull so the reader crosses the gap willingly.
Edge cases and failure modes
Empty pastes earn an informational verdict, not a failure. The checker reports that it found no labels and asks for a comma-separated list. This is by design: silence about missing input would read as approval.
Single-chapter pastes always pass under limited rules with zero switches and one POV. That verdict means nothing about the book; one data point cannot show a pattern. Never cite a single-chapter pass as evidence.
Two-chapter pastes with different heads score one switch across one gap, 100 percent, plus two singletons for two points: a warning from almost no information. Treat tiny-list verdicts as noise and check the full manuscript or a full act instead.
Long books past the sixty-label cap are judged on their first sixty chapters only, so split longer books into halves and check each separately. Near-duplicate labels split counts, so unify pairs like Jon and Jonathan before pasting.
Shared heads across timelines confuse the count honestly. If Young Maya and Old Maya are labeled differently, they count as two POVs with thin shares each; if labeled identically, they count as one head with a long stay. Choose the labeling that matches the promise you make to readers: different names on the chapter titles means different labels here.
Prologues and epilogues in foreign heads will flag as singletons. That flag is doing its job. Either the frame voice earns a second appearance or the frame should be retold by the lead. Do not exempt frames from the audit; readers do not exempt them from the experience.
Co-authored serials and round-robin drafts produce high-switch, high-singleton patterns naturally, since each author contributes a favorite head. Check the merged list early, agree on the surviving heads before anyone drafts act two, and reassign orphan chapters while feelings are still neutral.
Finally, remember what the checker cannot see. It never opens your manuscript, so within-chapter slips, where a limited chapter borrows one sentence of another mind, are invisible to it. It never judges quality, so a passing rotation of dull chapters passes. Pair the verdict with a craft read: this tool audits assignments, and your eyes audit sentences.
Genre-specific guidance
Romance readers attach to two heads and punish newcomers. Keep the leads above 40 percent each, hold singletons to zero, and treat the epilogue rule strictly: no new voice in the final chapter. Alternating every chapter is conventional here, so expect switch-rate warnings on two-lead alternation and keep them if re-anchors are strong.
Mystery and thriller depend on information control. A suspect or victim head hands the reader knowledge the detective lacks, which can deflate suspense. If a non-detective head is necessary, confine it to chapters whose information the detective could soon verify, and never let the combined non-lead share pass one third of the book. Singleton suspect chapters are the classic fault; expand, reassign, or cut them early.
Epic fantasy runs the largest casts, often four to six heads across long chapters. One POV per scene fits better than one per chapter here. Still enforce the two-chapter floor per head and watch the smallest share: a head below 10 percent of a long book reads as an interruption whatever the verdict says.
Horror trades on restriction. Single-head or tight two-head structures build dread because the reader cannot look away toward safety. New heads relieve pressure, which is the opposite of the genre job. Be stricter than the checker: treat even a passing three-head pattern as suspect unless each head is isolated and afraid.
Young adult favors voice intimacy and clean rules. One per chapter with two or occasionally three heads covers nearly the whole category. New writers in YA should treat any singleton as a cut and any warning as a rewrite signal, because the audience notices voice changes quickly and forgives them slowly.
Literary fiction tolerates more, including brief omniscient passages and single-visit heads used as deliberate instruments. The checker still flags these patterns, and it should: a deliberate break is still a break, and the flag is a prompt to confirm intention. Note each kept flag in the manuscript margin with its reason.
Science fiction varies by subgenre. Space opera behaves like epic fantasy with large casts and scene-level switching. Near-future thriller behaves like mystery with tight heads and strict information control. Match the rule to the subgenre you shelved toward, then hold the line the checker draws for that rule.
Across all genres, the smallest head carries the verdict. Before adding any new POV, check that the current smallest holds two chapters and real secrets. A cast grows by earning, one returning head at a time.
Plan against actuals, then draft the repairs
If you sketched POV assignments in advance with a POV plan builder, this checker is the other half of that pair: the plan records intent chapter by chapter, and this audit records what the draft actually does. Lay the two side by side after every major revision. A chapter where plan and audit disagree is a decision, not an error, until you decide: update the plan when the drift chapter reads better, revise the chapter when the plan was stronger. Teams and solo authors alike benefit from writing the decision down with the chapter number, because POV drift returns silently in later passes if the choice was only ever spoken.
Draft the repaired chapters in Co-Writer at /write, where each chapter file maps to one row of your matrix. Work down the flagged rows in order, rewriting each chapter inside a single head before moving on. Re-paste the updated label list here after each pass to confirm the verdict moved.
Questions writers ask next
Can a chapter switch heads mid-stream under one per chapter? No. That rule allows changes only at chapter breaks. A chapter that visits two minds needs either reassignment to one head or a rule change to one per scene with a real break on the page, and the break must be visible, not merely a paragraph gap.
Do prologues and epilogues count as chapters? Yes for this audit, because readers meet heads there. Label them in position and let the singleton flag speak. A frame voice that appears twice, prologue and epilogue, forms a legitimate two-visit head with a long stay between visits.
Is strict alternation between two heads head-hopping? Structurally it switches constantly, and the checker will say so with a warning. Artistically it is one of the most loved rhythms in commercial fiction when anchors are strong. Keep the rhythm, strengthen the pulls and anchors, and let the warning stand as a description rather than a condemnation.
How many POVs can a book carry? Two carries almost any commercial novel. Three works with clear share gaps, such as 50, 30, and 20 percent. Four or more demands scene-level switching rules and long stays. Above five, the audit usually fails under limited rules, and the draft usually feels that way too.
Should chapter titles announce the POV? Titles that name the head solve half the anchoring problem before the first line, and they make your pasted list trivially easy to build accurately. If your design hides heads for suspense, keep the titles coy but pay for it with stronger opening paragraphs in each new-head chapter.