Verified against ChatGPT · 2026-08-09
Build a branching interactive story where choices actually diverge instead of reconverging on the same line
Produces a choose-your-own-path story with a mapped decision tree, where early choices meaningfully change what the reader experiences rather than all funneling back to one identical scene, with the branch map shown alongside the text.
The prompt
Ready to copy — highlighted parts are example details you can swap.
Write a branching, choose-your-own-path interactive story with the structure below. The most common failure in generated interactive fiction is that branches look different for one paragraph and then reconverge on the same scene regardless of what the reader picked — I want the choices to actually matter for longer than that. OPENING SITUATION The reader wakes up on a supply ship that's just received a distress signal from a research station that officially stopped transmitting six months ago. FIRST DECISION POINT Investigate the signal immediately alone, report it to the captain and wait for orders, or quietly ignore it and continue the ship's original course. DESIRED BRANCH DEPTH At least two further decision points after the one where the choice was made. ENDING VARIETY Four endings: one where the reader escapes alone, one where they escape with a rescued survivor, one where they're stranded, and one where they choose to stay. RULES Start with the opening situation and present the first decision as two or three genuinely different options, each implying a different kind of story (different risk, different relationship, different tone) rather than two paths that are cosmetically different but functionally the same scene. Carry the consequences of each choice forward for at least At least two further decision points after the one where the choice was made. — a choice's effects should still be visible or relevant that many decision points later, whether as a changed relationship, a resource gained or lost, or a door now closed that was open before. Before writing the full text, output a branch map showing the tree structure (which choice leads to which node, and where nodes do or don't reconverge) so the structure is checkable before reading the prose. If two branches do reconverge at some point for practical length reasons, make that convergence a deliberate, acknowledged choice you flag in the map — never let branches accidentally collapse into an identical scene without noting it as a decision. WHAT NOT TO DO Do not write a decision point where both options are followed by essentially the same next paragraph with a few swapped nouns — if that's the only way to keep the branch manageable, say so honestly rather than disguising it as a meaningful choice. Do not make every path funnel to one single ending; produce the number of distinct endings specified in Four endings: one where the reader escapes alone, one where they escape with a rescued survivor, one where they're stranded, and one where they choose to stay., each reflecting a different accumulated set of choices, not just a different final sentence bolted onto the same outcome. OUTPUT FORMAT 1. The branch map, as an indented outline or simple tree diagram in text. 2. The full story text, with each node labeled to match the map, and choices presented clearly at the end of each node. 3. One paragraph noting any place branches were deliberately merged for length, and why.
Customize
Optional — swap in your own details for the highlighted parts above.
Why this works
Interactive fiction generated without explicit structural constraint tends to produce what's sometimes called an illusion of choice — branches that diverge cosmetically for a paragraph and then reconverge on an identical scene, because maintaining genuinely separate story states across multiple future decision points is combinatorially expensive, and a model optimizing for a complete-feeling response within a reasonable length will quietly collapse branches back together rather than track divergent state all the way through. Requiring a branch map before the prose forces GPT-5.1 to commit to the tree's actual shape as a discrete planning step, which makes silent reconvergence visible and checkable rather than something that happens invisibly inside prose generation — a map is much harder to fudge than a paragraph transition. Explicitly requiring consequences to remain visible a minimum number of decision points later targets the specific failure of a choice mattering for exactly one immediate beat and then evaporating, which is the most common way "branching" fiction ends up not actually being branching in any way a reader would notice on a replay. Permitting deliberate, flagged convergence rather than banning it outright is important because real interactive fiction does converge sometimes for practical reasons — the instruction isn't that convergence is always wrong, it's that unflagged, accidental convergence disguised as meaningful choice is the actual defect, and making the model own that trade-off explicitly prevents it from being hidden inside the narrative. Requiring a specific number of genuinely distinct endings, each tied to an accumulated choice history rather than a single final swapped sentence, closes the same gap at the story's terminal nodes that the branch-depth rule closes in the middle.
What you get back
Branch map: - Investigate alone -> discover the signal is a recorded loop -> [go deeper / turn back] - Report to captain -> squad sent together -> [reader leads / reader follows] - Ignore signal -> station's automated distress becomes ship's problem later -> forced convergence, flagged below
Verified against
ChatGPT GPT-5.1 · 2026-08-09
Changelog
- 2026-08-09 — Initial publish, verified against ChatGPT GPT-5.1.
Need this built into your business?
If a prompt isn't enough — what Scult builds, built and maintained for you — that's Scult's day job.
EXPLORE WHAT SCULT BUILDS
