One next move per engine, for the answer you want
The win plan takes one prompt you are losing and gives each engine one thing to do. It shows up to five sources that decided the answer, whether you are on them, your rung (proven, better or new), the next action with a credit estimate, and the day the answer is read again. Engines are never averaged.
7 days free · card required
ChatGPT · rung: new · n = 57Nothing of ours is among the sources that decide this answer. They are technologyadvice.com, trylyra.ai and axissuite.ai roundups. About 8 credits.
Perplexity · rung: new · n = 52getapp.com is what the answer was built from, beside softwareadvice.com and seoptimer.com. About 4 credits.
- Read again
- after the first action lands
No change proven yet: this is work to do, not a result.
The win plan for a prompt attensira.com loses. Estimates are as the product gave them; the meter on the run is what is charged.
Chief of staffattensira.com workspace, read over the Attensira MCP · - Chief of staff · Works from the plan every morning and hands each action to the role that does it.
Researcher · Reads the sources that decided the answer and what changed on them.
How is the win plan built?
It is composed over the deciding-sources map, so the map, the prompt page and the tool can never disagree about which page decided an answer.
Read who decided the answer
Up to five deciding sources per engine, each with its share, its n, and whether you are already on it.
Place you on a rung
Proven: you are cited among the deciding pages. Better: you are cited and a competitor sits above you. New: nothing of yours is there yet.
Pick one action
Fix a page, get listed, pitch a listing, answer a thread, or write something citable that does not exist yet, with the page to work on and the citations it rests on.
Book the re-read
Every plan ends with the day the answer is read again. If nothing is booked yet, it says “after the first action lands” instead of naming a date.
What does a lost answer look like?
Here is one we lose. Perplexity answers how to block AI crawlers from ten other sites' guides and names us nowhere. The win plan starts from answers like this one, per engine.
Question: how to block AI crawlers in robots.txt
Perplexity answered: To block AI crawlers in robots.txt, add a dedicated User-agent group for each crawler and Disallow: / (or the specific paths you want blocked). Example: User-agent: GPTBot Disallow: / User-agent: ClaudeBot Disallow: / User-agent: Google-Extended Disallow: / (Repeat for other crawlers as needed.) Note: blocking robots.txt may not prevent indexing; for private content, use authentication, and use noindex if you don’t want pages appearing in search results.
Sources cited
- 1geotoolbox.ai
- 2scalableloops.ai
- 3tooltrace.io
- 4maxaeo.ai
- 5centium.ai
- 6qwairy.co
- +4 more
Not namedattensira.com is missing from this Perplexity answer.
What does each part of the plan mean?
Every line is built from stored answers, and each says how much evidence stands behind it.
| Deciding sources | Up to five pages per engine that decided the answer, ranked by citation share × position, each with its n and whether you are on it. |
|---|---|
| Rung | Proven: you are cited among the deciding pages. Better: you are cited and a competitor sits above you. New: nothing of yours is there yet. |
| The action | One of: fix a page, write a new page, get listed, pitch, or answer a thread. Each names its target and the citations it rests on. |
| Evidence | Measured at 7 or more answers, published at 2 to 6, hypothesis at one. Until 20 reads of a source exist, the move comes from the per-engine table and says so. |
| Credits | An estimate per action and for the whole plan, so you can see the cost before the crew starts. The run's meter is the bill. |
| Per engine | Each engine gets its own sources, rung, action and the day it is read again. Engines are never averaged. |
What will the win plan not say?
It names the evidence and one move. It never names an outcome.
- No forecast: no predicted rank, no predicted lift, no percentage that has not been measured.
- No movement words until a re-read comes back moved.
- One action per engine, never a list of five.
- The credit estimate is an estimate. The meter on the run is what you are charged.
- Every number carries its evidence rung: measured at 7 or more answers, published at 2 to 6, hypothesis at one.
And what does the crew ship to win it?
Our own repository is not connected, so attensira.com has no pull request to show yet. This is a separate example of the same move for Ledgerly, a fictional brand: the plan's first step as a change you can review, an answer passage at the top of the guide the question is about. It is a draft until you merge it.
Answer “How do I block AI crawlers in robots.txt?” first
attensira2 checks passed
content/guides/robots-txt.mdx
- unchanged: ## Blocking AI crawlers
- added: Add one User-agent group per crawler, then Disallow: /. Blocking a search crawler can keep your pages out of that engine's answers.
Web engineer · Opened the fixExample · Ledgerly, a fictional brandWhat else do people ask about the win plan?
Where do I find the win plan in Attensira?
Why one action per engine instead of one plan?
Does the win plan predict results?
Does the crew act on the plan without me?
Which prompt should I pick first?
Pick a prompt you are losing. Get the next move.
7 days free · card required