Prompts

Give me a week-over-week pulse on how this brand does in AI answers.

If this organization runs more than one project, call list_projects() first. Then ask
which project the user means. To cover every project instead, repeat all the steps below
once per project, with its id as `project`. Wait a few seconds between projects, so that
one burst does not spend the organization's per-minute allowance. A key or a connection
scoped to one project needs none of this.

Call get_digest() first. It returns the server's week-over-week breakdown of the last
full calendar week against the week before. Then call get_overview(period="7d") and
get_perception(period="7d") for the current splits and the perception detail. The
digest's window is the calendar week. The windows of the other two end at the newest
collected day. Put the digest's `week` span in the headline. Do not reconcile small
differences between the two. They are different windows, not a discrepancy.

Report in this order:

1. The headline. Report `visibility` against `visibility_previous` from get_digest, over
   `answers` analyzed answers, for the `week` span. If `previous` is null, report the
   prompt figures as standing numbers and never as a delta. Then, if `prior_other` is
   true, report that the prompts of the prior week differ, so the prompt figures have no
   comparison. Otherwise the prior week has no measurement, and the comparison starts
   next week. If `measured` is false, no answers were analyzed in the last full week.
   Report that, and report get_overview's trailing window instead.

2. The sharpest change. When `highlight` is present, it names one platform and one
   prompt. That platform named this brand first in `first` of its `answers` to that
   prompt last week, and in `first_previous` answers the week before. Start with it, and
   quote the prompt text.

3. Movers. Each entry in `movers` holds a `movement` verdict. `appeared` and
   `disappeared` are events of their own: mentions started or stopped for a whole week.
   `up` and `down` come with both values. Report each mover.

4. Ground shifts. Each `shifts` row passed a share-of-voice reporting floor. Report each
   brand's share_of_voice against share_of_voice_previous. Report `position` against
   `position_previous` where it moved. A lower position is better. `flipped` lists
   prompts that started or stopped naming this brand. Quote one or two.

5. Sources and standing issues. `domains` were newly cited last week. `actions` are the
   Actions first built last week, each with its headline and its answers. When
   `previous` is null, they are the week's most cited domains and the top of the
   worklist, so never call them new. Report `crawlers_blocked` or `prompts_silent`
   whenever either is non-zero. Both are standing problems, not last week's news. For a
   blocked crawler, call get_crawlability() and quote the rule and its line. If
   get_ai_visits() has a connection, report the sessions AI assistants sent last week.
   If get_crawler_visits() has a window, report any crawler whose errors rose.

6. How the answers speak about this brand. From get_overview, report `sentiment` as a
   split, not as an average. Report any negative count on its own, however small. Report
   `tiles.position` with its `positioned_answers`, and say that a lower place is better.

7. What this brand is known for. get_perception's `rows` are attribute and objection
   themes per brand. Each theme has a 0-100 score and verbatim quotes behind it. Start
   with this brand's top attribute and its strongest objection. Report any objection
   that a competitor scores worse on. That is sales material. If `entitled` is false,
   Perception is not part of this plan. Report that, and do not report "nothing found".
   The digest's `perception` block shows whether new perception answers were collected
   last week. Start with any `first_objections`. Word them as first seen in the stored
   answers, not as the platforms changing their minds. When `prior` is false, they are
   the whole list. Do not call them new.

End with one paragraph: what changed, and the single thing worth doing about it this
week.

Find the competitors that moved, and report whether the movement matters.

Call list_competitors(period="7d") and get_overview(period="7d").

Treat a brand as moving when its `delta` differs from zero by more than 10 percentage
points. `visibility` and `delta` are fractions, so compare against 10/100.

For each mover:
- Report its name.
- Report its before and after values as percentages.
- Find where it moved. Call get_overview(platform=..., period="7d") for each platform in
  `per_platform`.

Then read `detected`: brands that the analysis found in answers and that nobody
monitors. A new brand there is the earliest signal available. It counts toward no score
until somebody promotes it. If `emerging` is false, that list is not part of this
organization's plan. Report that. Do not report an empty list as "no new competitors".

Report no movement for a brand whose `delta` is null. That is missing history, not a
flat line.

Call list_products(period="7d"). When `totals.shopping_answers` is above zero, report
the brands whose products take the first card, with `first_card` out of
`totals.shopping_answers`. FullReach AI records product cards only from ChatGPT and
Google AI Mode.

Decide which AI platform deserves attention next.

Call get_overview(period="7d") for the whole picture. Then call it once more for each
platform in `per_platform`, with platform= set each time. Every platform then gets its
own visibility, Share of Voice, sentiment and competitor table.

Build one table with these columns:
- platform
- answers collected
- answers that name this brand or any monitored competitor (`monitored_answers`)
- this brand's visibility
- its Share of Voice
- the strongest competitor on that platform.

Two readings look identical in a visibility score but mean opposite things. Keep them
apart. Report which reading applies to each platform:

- Low visibility, high `monitored_answers`. The platform answers this category by naming
  companies, and it names other companies. This is a losing position, and a winnable
  one.
- Low visibility, low `monitored_answers`. The platform answers without naming any
  company. There is nothing to win there yet. Work aimed at it is wasted.

End by reporting the platform to work on, and why. If the honest answer is "none of them
yet", report that. Do not pick the least bad one.

Grade the prompts, and report which ones to change.

Call list_prompts(period="7d"). Sort every row into four groups:

- **Winning**: high `visibility`. Most answers name this brand.
- **Losing**: low `visibility` but high `monitored_answers`. The answers name companies,
  but not this brand. This is the group worth working on.
- **Silent**: `silent` is true. The recent answers named no company at all, so the
  prompt is not a buying question. Recommend rewording or dropping these prompts. No
  amount of work on this brand moves them.
- **Unproven**: few `answers`, or `failures` are a large share of them. These rows are
  not a result yet, and not evidence of anything.

For the two or three worst rows in **Losing**:
1. Call list_prompt_answers(prompt_id=...).
2. Call get_answer(answer_id=...) on one recent answer.
3. Quote the sentence that names the competitor.

The quote is the finding. The score was only how you found it.

Call list_suggestions() as well. Each suggestion is a replacement prompt that the
product already derived from this project's own answers. Each one holds the evidence it
came from. Where a suggestion covers the same ground, prefer it to wording that you
invent. An empty list means that nobody has generated suggestions yet. Report that, and
do not fill the gap yourself. Generating is a button in the app.

Give each group a count. Then report the three prompts most worth changing, and what to
change them to.

Compare how this brand does across the project's markets.

Call list_markets(period="7d").

`markets` has one row per market. `tracked` is false on a market that has
answers in the period but no active prompt now.
The rows hold **counts, never rates**:
- `prompts`: how many active prompts run there
- `answers`: how many answers were collected there
- `named`: how many named this brand or any monitored competitor
- `present`: how many named this brand.

Visibility is `present / answers`. Calculate it. Do not look for it in the result.

If only one market has `tracked` true, report that and stop. A comparison of one
market against itself is not a finding. A market recorded as ZZ is Worldwide:
its prompts steer at no country at all.

Each market runs its own prompts, which may differ in wording and language.
Say so when the markets' prompt counts differ, because two markets asked
different questions compare the questions as much as the markets.

Otherwise, report for each market:
- its visibility
- how it differs from the strongest market
- whether the difference is a losing position or an empty category.

Where `named` is far below `answers`, the answers in that market name nobody. That
is a different problem from answers that name a competitor. To confirm the
worst market, call get_overview(market=..., period="7d"). Read which
brands the answers name there instead.

Then split the same window by who asks. Call list_personas(period="7d").
`personas` has one row per persona, the last row being the prompts with no
persona, with the same counts. A persona with a `sentence` is asked as that
buyer: its sentence is sent in front of every prompt attached to it. Where one
persona's visibility is far below the others', the answers to that buyer name
somebody else, and the prompts attached to it say who was asked. One persona
is no comparison: report it and move on.

Then split it by subject. Call list_topics(period="7d"). `topics` has one
row per topic, the last row being the prompts with no topic, with the same
counts. A topic far below the others is a subject where the answers name
somebody else. One topic is no comparison: report it and move on.

End with the one market worth attention. If the difference does not justify
splitting effort, end with "the difference does not justify splitting effort".

Find where this brand is strong and where nobody sees it, by topic and platform.

Call get_breakdown(rows="topic", columns="platform", period="7d"). In a cell, this
brand's visibility is present/answers. Its visibility over the whole grid is
totals.present/totals.answers. Call that the baseline.

Build a table with one row per entry in `rows.values` and one column per entry in
`columns.values`. Write each cell's visibility as a percentage, with its `answers`
beside it. Leave a pair with no entry in `cells` blank: no answers were collected there.

Judge a cell only when it holds at least 10 answers. Name it when its visibility differs
from the baseline by 10 points or more:
- below the baseline: a blind spot
- above the baseline: a strong position.
Say how many cells hold fewer than 10 answers. They are too small to judge.

A blind spot with a high named/answers is winnable: the answers there name companies,
but not this brand. A blind spot with a low named/answers is not: the answers name no
company at all. Report which kind each blind spot is.

For the two worst blind spots, call list_prompts(period="7d") with topic= set to the
row's id and platform= set to the column's id. Pass "none" for a null topic id. Report
the prompts there that this brand loses.

Audit which websites supply the answers in this category, and where this brand's own
site stands among them.

Call list_sources(period="7d"). Report the ten most cited domains, each with its
citations as a share of `totals.citations`. `brand_domains` gives the rank of this
brand's website and of each monitored competitor's website. Report those ranks too. For
each top domain, report the `kind` of its listed URLs: a press site can carry a roundup.
Report a null kind as not labeled yet, and never guess one.

Then call list_sources once for each platform in `per_platform`, with platform= set each
time. Report where the platforms disagree: a domain that one platform cites often and
another never cites.

Call list_source_urls(period="7d") to read the URLs of the top domains past the URLs
that list_sources lists. Pass `next_offset` as `offset` only until the rows leave the
top domains. Count the URLs of each `kind`, and name the most cited roundups and
directories: a company can work to appear on those pages. If the tool refuses for the
plan, use the URLs that list_sources returned, with their `kind`, and say that they are
the top rows only.

Call list_own_pages(period="7d"). Report:
- the own pages that the answers cite most
- pages with `read_answers` but few `cited_answers`: a platform read the page and cited
  something else. This holds only for the platforms in `reading_platforms`.
- pages with `crawler_visits` or `ai_sessions` but no `cited_answers`: AI crawlers or
  visitors reach the page, and no answer cites it
- pages with `unnamed_answers`: an answer cites the page and does not name this brand.

These tools count citations. They do not read the pages. Never describe what a page
says. End with the three sources most worth work, each one tied to a row that you
reported. list_actions() returns the worklist that the product built from the same
citations. Where an Action's `target` is the same source, name it.

Measure what changed in AI answers after the launch on <L>.

Ask the user which day the campaign launched, written as YYYY-MM-DD. Do not guess it,
and do not take it from the answers. Call it L below.

Call get_overview() first. Its `window.end` is the newest collected day. Count the days
from <L> to that day, both included, and call the count N. Use at most 90. If the launch
is after `window.end`, report that no answers were collected after the launch, and stop.

Call get_overview(period="<N>d", until="<the day before L>") for the window before the
launch, and get_overview(period="<N>d") for the window after it. Compare this brand's
visibility, Share of Voice, sentiment and position between the two. Put each window's
`window` span and `totals.analysed_answers` beside its numbers. A read's own `delta`
compares with the window before that read, which is not always the other window. Compare
the two reads directly.

Call list_competitors and list_sources with the same two argument sets. Report which
competitors gained or lost the most, and which cited domains appear only after the
launch.

At one answer per prompt, platform and market a day, a change of a few points can be
sampling noise. Report a change as a result only when both windows hold enough answers
to carry it, and say how many they hold. If nothing moved beyond that noise, report that
the launch shows no measurable change yet.

Plan this week's work to get this brand into more answers.

Call list_actions().

`actions` is the worklist, in `position` order. Plan from this week's list: the open
rows whose `list_week` equals the response's `list_week`. Each has a `headline` that
words the verb, and `targets` counts the losing answers no Action above it targets.
Leave out `done` and `dismissed` rows, and any row with `faded_at` set.

Read each kind for what it asks:
- `unblock`: the site's robots.txt turns away a search crawler, or a crawler that
  fetches a page when a person asks. It is the cheapest fix and the one with the
  strongest evidence. Say it first.
- `logs`: the Cloudflare logs record no request to this host. Until a developer fixes
  that, nobody can see whether AI crawlers fetch its pages.
- `sitemap`: no AI crawler fetched the own pages the Publish Actions improve. The action
  is to submit the sitemap to Bing and Google.
- `gate`: a register, association or ranking. Inclusion is by application.
- `list`: a publisher of best-of lists. The action is an editorial pitch.
- `marketplace`: a review site. The action is a profile this brand creates.
- `press`: a publication. The action is a pitch to its editors.
- `platform`: a place this brand publishes under its own name.
- `community`: a forum. Taking part is earned and slow. Say so.
- `publish`: a prompt whose answers cite tool makers' own lists. Its target is the
  prompt's id. The action is a page of this brand's own that answers it, or an edit of
  the own page its `headline` names. Its `evidence` is the brief: the questions, the
  `phrase` the engines searched most, the competitors to name and the lists the answers
  cite instead.
- `rival`: a retired kind. A done row is history.
- `vendor`: a list a tool maker wrote about its own tool. It never goes on the list.
  Answer it on a page of this brand's own, never by asking the maker.
- `other`: the kind could not be read. Open one page before you judge it.
- `opening`: an objection the answers raise against competitors. `proposed` means the
  customer has not confirmed this brand does better there. Plan no work on a proposed
  Opening.

**Size it before ranking it.** `covered_answers` of `lost_answers` is how much of the
problem the destinations explain.

**State the claim exactly.** The answers that cited a page named those competitors. This
data cannot show whether a competitor's name appears on the page. So write "the answers
citing this page named X". Never write "this page mentions X".

For the top three Actions, call get_answer(answer_id=...) on one of the evidence's
`answer_ids` and read what the citation supported. Report the `searches` in the
evidence: they are the words the platform searched before it cited the page. Only
ChatGPT reports them. list_searches() holds every search in the window.

Call list_sources() and compare `totals.citations` with `totals.answers`. In a category
whose answers cite little, getting written about moves less than getting mentioned does.
Read the worklist with that in mind.

Call list_own_pages(). `unnamed_answers` counts answers that cite the brand's own page
and still do not name the brand. Report those pages beside the worklist, as pages the
answers already trust.

End with the three Actions most worth this week, each with its headline, its answers and
one sentence on why.

Tune this project's prompts and competitors, with my approval.

If this organization runs more than one project, call list_projects() first. Then ask
which project I mean.

Read before you propose anything:
1. Call list_suggestions(). Each suggestion comes from this project's own answers, with
   its evidence.
2. Call list_prompts(period="7d"). A row with `silent` true is a prompt whose recent
   answers named no company at all.
3. Call list_competitors(period="7d"). `detected` holds brands the answers name that
   nobody monitors. A row's `proposals` are other spellings of a monitored competitor
   that the answers used.

Then propose one plan, as a numbered list grouped by tool:
- accept_suggestions(suggestions=...) for the suggestions worth asking. For a `silent`
  row with `replaces`, say whether to pause the prompt it replaces.
- dismiss_suggestions(suggestion_ids=...) for the ones that are not buying questions.
  Say that a dismissal is permanent.
- reword_prompts(prompts=...) or pause_prompts(prompt_ids=...) for silent prompts.
  Prefer a suggestion that covers the same ground to wording that you invent.
- resume_prompts(prompt_ids=...) for paused prompts that I ask for again.
- add_prompts(prompts=...) for questions that I name. Do not invent prompts that I did
  not ask for.
- set_markets(prompt_ids=..., start=..., stop=...) for prompts that I want asked in
  other markets.
- rename_topics(topics=...) or set_topic(prompt_ids=..., topic_id=...) only for topics
  that I ask to relabel or prompts that I ask to regroup.
- add_competitors(competitors=...) for detected brands that are real competitors. Each
  needs its website. Ask me for any website that you do not know. Never guess one.
- set_competitor_aliases(competitors=...) for the proposals that are spellings of the
  same company. The tool replaces the whole list, so pass the row's current `aliases`
  with the approved proposals. A namesake is a different company. Leave it out.
- stop_monitoring_competitors(competitor_ids=...) for monitored competitors that I name.

Say what each group costs: each accepted suggestion takes one prompt of the plan's
allowance. An added prompt takes one for each market it runs in. A resumed prompt takes
one for each market it restarts, and a started market takes one. Then stop and wait for
my answer. Make no write before I say yes. Write only what I approve.

Each write tool is all or nothing. When one refuses, report its refusal and ask how to
change the plan. Do not retry with items left out. End with what changed, and say that
new prompts get their first answers at the next daily collection.