A clear question. A traceable answer.
On this page
One offering. One buying category. One buyer market.
A Decision Loop follows one buying decision: one offering in one buying category, for buyers in one country. That country is your Buyer Market. A company that sells in two markets runs two loops.
Each loop runs the whole way round: the Decision Map, the Work Plan, the Action Pack, your work, then Measure Again and the Before/After Snapshot. It is not a company-wide visibility score.
What you tell us.
You set up your study after you sign in.
- WebsiteCitePack reads it and suggests the offerings and buying categories it could study.
- Buyer MarketThe country your buyers are in. It starts at the United States, and you can choose any other.
- Offering and buying categoryPick from the suggestions, or type your own.
- Your own wordsWhat the offering is best at and who it’s for. If you like, add who it isn’t for and facts you stand behind, such as a price, a limit or a certification. CitePack keeps your exact wording.
- Five questionsA buying situation for each one, with up to two buyer specifics.
Then you review the study and pay. Next comes the part that decides what gets asked: your questions.
Five buying situations.
Buyers ask different things at different points in a decision. Each of your five questions covers one buying situation, and the five together cover the whole decision.
- Category considerationThe buyer is just starting to look. Shows who is considered at all. Usually left broad, with no specifics.
- Job suitabilityThe buyer describes the work they need done. Shows who is named as right for that job.
- Mandatory eligibilityThe buyer has a condition they won’t bend on. Shows who is still named once it applies.
- Balancing prioritiesOne thing matters more to the buyer than another, such as price over reporting depth. Shows who is named once that priority is set.
- Change of solutionThe buyer is moving away from what they use now. Shows who is named as the next move.
A question can carry up to two buyer specifics: the work to be done, a must-have, a preference or the buyer’s starting point. CitePack suggests specifics from your website, and you can write your own. Nothing is added just to fill a slot.
CitePack drafts the five questions. You read each one, change any wording you like and approve all five together. Once approved, the questions are sealed. Every engine gets them exactly as written, and they never change for that study.
Four engines, three repeats each.
- EnginesChatGPT, Claude, Gemini and Perplexity, each through its official API.
- ModelOne fixed model per engine, matched to the model its free app gives buyers. Every answer records the model that wrote it.
- Web searchOn for every engine, so answers can draw on pages from the web.
- Clean repeatsNo account history or memory shapes an answer. Every repeat uses the same question, model, market and settings.
- RepeatsEach question three times on each engine. 5 questions × 4 engines × 3 repeats = 60 observations.
Why three times? One answer can’t tell you whether the next will match it. Three repeats on every engine show where an answer holds and where it shifts. Every repeat is kept and shown.
In the app, each buyer’s screen is shaped by their own history and settings, so one screenshot describes one person. A Decision Map shows the pattern across 60 controlled observations in your buyers’ market.
5 questions × 4 engines × 3 repeats. Every answer is read and kept in the map.
ChatGPT
Claude
Gemini
PerplexityQuestion 1Your market, on every engine.
By default, your Buyer Market reaches every engine on every repeat, two ways at once. Each question names it, and each engine is set to it. We call this controlled market execution.
- ChatGPTIts own location setting, set to your country.
- PerplexityIts own location setting, set to your country.
- ClaudeIts own location setting, set to your country.
- GeminiA location point in your country.
CitePack drafts your questions with your market in them, and you approve every word. After that, your approved question is never changed.
You can turn this off before you approve. Then the questions don’t name your market and CitePack sends no location to any engine, so each one answers from its own default. It’s your call, and we recommend keeping it on.
Your map records, for each engine, how the market was applied.
What’s in your map.
- OverviewHow often you were put forward across all 60 observations, how often you were named first, and how often no option clearly came first.
- Questions 1 to 5A page per question, with every answer, the order it names options in and what it says about you.
- AI enginesHow each engine built its answers. What it searched, whether it used your site, how it laid the answer out, and how early and how much it wrote about you.
- Who gets pickedEvery other name in the answers, who comes before or after you, and who is listed first.
- What AI saysThe headline phrases on your pages next to the answers’ words about you, how the answers wrote your name, and the prices, free plans and amounts they stated, checked against your website.
- Evidence & TrustThe study as it ran, and every answer exactly as it came back.
Every number shows what it’s counted out of, and opens to the answers behind it.
Keep the differences in view.
Being put forward as an option, being mentioned and being cited are different things, and the map keeps them apart. An answer can mention a company without offering it. It can cite a page without recommending that page’s product.
When an answer lists options in a clear order, the map records your place in it: first, second, third and so on. If the order isn’t clear, the map says so instead of guessing. Every other name is read the same way. Who gets picked shows where each one sits against you, counted only in the answers where that order can be read.
A place in an answer is what that answer said. If one answer picks a single option, that pick belongs to that answer alone. It never becomes an overall verdict.
Each voice keeps its source.
What you confirm about your business, what your published pages say and what an AI answer says are kept separate. You can read them side by side, and the map never passes one off as another.
CitePack reading your website is not the same as an engine citing a page, or finding one through search. Cited pages and pages found by search are kept apart. A source record shows which page an answer cited, or which page its search found. It does not explain why the answer chose its wording.
Behind every finding sit the answers themselves. Each one can be read in full, with its engine, model and date.
See how it looks: Evidence & Trust in the example map
The Work Plan, beside your map.
The Work Plan has its own address, linked from your map. It follows what the engines did: the searches they ran, the companies they looked up by name, the pages they read. CitePack reads those pages too and shows where you and the company listed first sit on each one.
Then it ranks the work. Each item names the question, the answers and the pages or words that show it, how big the job is, who does it, and the check to run the next time you measure. If no answer points to a job, it isn’t in the plan.
Where another company comes first and no page of yours was used, you get the outline of the page to build: a title with the year, the engines’ own searches, the columns shared by the lists they used, and an FAQ. Facts about you are quoted from your site, the Action Pack carries the draft, and nothing about a rival is stated for you.
The Action Pack, beside every item.
Every Work Plan item comes with the material to do it, grouped by kind. Open a piece and it is ready to use or send.
- Email draftsThe pitch or the updated facts for each page the engines used.
- Page draftsThe outline of each page to build, from the engines’ own searches.
- RewritesThe one claim you want the answers to repeat, in every place it belongs.
- Correction notesOne line for each fact the answers state differently from your site.
- Technical fixesWhat stops the AI crawlers reading a page, and what to change.
Facts about you are quoted from your site. Nothing about another company is stated for you. Do the work and mark each item done.
Measure Again, then the Before/After Snapshot.
Measure the same decision again on Day 28 or X, X being the day you choose. We recommend Day 28, 28 days after your first map, and the choice is yours. The same sealed questions go to the same four engines, three times each, in the same buyer market. Only time has passed, so what differs is what the engines now say.
- Day 0 beside Day 28 or XEvery answer from both measurements, side by side. Day 28 is the day we recommend; X is any other day you pick.
- Each engineMarked Same or Different, with how many of its fifteen answers changed. Where an engine’s model changed between the two measurements, the snapshot says so beside that engine.
- Each item’s checkEvery Work Plan item carries a question, answered on the new map beside what you marked. An item you did not do shows as not done.
- The Before/After SnapshotWhat moved, or that nothing did. A map that shows no change is a result too.
The new map comes with its own Work Plan and Action Pack, so one loop holds 120 answers, two Decision Maps, two Work Plans, two Action Packs and the Before/After Snapshot between them.
Your account home sits above your decisions. Across several, it shows each one’s latest standing and what moved since its previous map.
One Measure Again is included in every loop. For another measurement of the same decision, or a monthly rhythm, email us.
From approval to your map.
After you approve your questions, CitePack asks them, reads every answer and builds your map. It reaches your account in 30 minutes on average, with the Work Plan and the Action Pack beside it. We email you when it’s ready. If our checks find a mistake, we hold the map and put it right first. If it isn’t delivered within 24 hours, you get a full refund.
Only you can open your map.
What you can rely on.
Every finding leads back to the answers behind it, and you can open each one. Where answers differ, the map shows it. Where the evidence runs out, the map says so.
We don’t sell outcomes. No visibility score, no generic list of fixes, no promise of citations. Every Work Plan item and every Action Pack piece points to the answers and pages behind it. Measure Again reports what changed, or that nothing did, and nothing here claims to know an engine’s reasoning or that a change caused a move.
Know what you’re looking at.
Crosstrow is the illustrative offering we use to show how a loop reads: its map, Work Plan, Action Pack and Measure Again. Its figures are not measured engine results, customer outcomes or a sample study.
The full example map is an illustrative example as well. Crosstrow and the companies its answers name are illustrative, and its answers and figures are not study results. Its pages work the way a real map’s do.
The founding cohort is open: ten seats, first come, first served.