create_brief
ChatGPTCreate a new content brief and start the research + strategy pipeline (SERP analysis, persona synthesis, outline generation, title suggestions). Consumes BRIEF_GENERATION credits. Returns a briefId to poll with get_brief_progress. Use create_blank_brief instead if the user wants to fill the brief manually.
finalize_document
ChatGPTMark the outline document as ready. After finalize, the brief can proceed to content generation via trigger_content_generation.
get_article
ChatGPTFetch the generated article for a brief — markdown content, section breakdown, metadata, and internal review notes. Internal authoring markers are auto-stripped before delivery, and the article is checked against the brief outline; when it fails any compliance check its status is needs_review (still readable, but a human should resolve the flagged gaps before publishing).
get_brief
ChatGPTFetch full details for a single content brief, including pipeline outputs (research, strategy, title suggestions).
get_brief_progress
ChatGPTPoll pipeline progress for a brief. Returns currentStep / totalSteps / progress percentage and any errorMessage.
get_document
ChatGPTFetch the structured outline document for a brief — metadata, keywords, strategy, and sections with topics/instructions.
get_domain
ChatGPTFetch full details for a single domain: brand attributes, aliases, tracked markets/languages, owned sources.
get_domain_alignment_aggregated
ChatGPTBrand-vs-content alignment scores aggregated over a date range. Surfaces how LLM depictions compare with the brand's stated attributes. modelBreakdown lists each model with its own average score (one report is produced per model per scan); modelAttributeScores splits it by attribute id. attributeBreakdown names each attribute id; removed: true marks an attribute since dropped from the brand profile, kept for its history. Citations are PAGINATED: citations.items holds ONE page (default 20 rows, max 100), with citations.totalItems/hasMore reporting the rest — the full list runs to tens of thousands of rows and does not fit in a single response. Move through it with citationsPage, or pass citationsPageSize: 0 for the scores alone.
get_domain_explorer_aggregated
ChatGPTFree-form entity/mention explorer: aggregated topics and entities LLMs associate with the brand. Answers "what is my brand known for in LLM answers?". Returns topKeywords and topSources (each item has count = integer and percentage = 0-100, may be fractional), summary (integer counts), domainSourceAnalysis, availableModels, availableTopics, reportCount, dateRange, totalAnswers. All percentages are 0-100 — present them with "%". Optional topicId narrows the aggregation to one topic (visibility + competition only; sentiment/alignment are domain-level and excluded). NOTE: the raw per-search citation list is omitted to keep the response compact — use list_explorer_citations for individual citations.
get_domain_report
ChatGPTFetch one bi-weekly AI visibility report as it was stored when built: the domain reading on the last day of the two weeks and its verdict against the first day (too-close, up or down, read on the 95% ranges), every reported topic ranked with its verdict, daily series, Shifts, tracked competitors, share of voice, its most cited source URLs (core sources, or the top other sources when the two weeks have none) with the URL and citation totals of that group, and content gaps, and the topics left out with the day their first score comes.
get_domain_sentiment_aggregated
ChatGPTTop-line "how are LLMs describing my brand?" answer: aggregated sentiment scores, quotes, and trend over a date range. modelBreakdown lists each model with its own average score (one report is produced per model per scan). Citations are PAGINATED: citations.items holds ONE page (default 20 rows, max 100), with citations.totalItems/hasMore reporting the rest — the full list runs to tens of thousands of rows and does not fit in a single response. Move through it with citationsPage, or pass citationsPageSize: 0 for the scores alone.
get_generation_status
ChatGPTPoll content generation progress after trigger_content_generation. Shows current phase, progress %, and any errors.
get_topic
ChatGPTFetch a single tracked topic by ID with its configuration (prompts, LLM models, markets). name is the stored label; displayName is what the topic is shown as and what its competition prompts compare brands on, and competitionSubject overrides that subject when set. Both are absent when unset, in which case name is used.
get_topic_competition
ChatGPTRecent competition runs for a topic, newest first: the latest 10 analyses, each with its reportDate, YOUR brand averageScore (0-100 competition score), scoreVariation and availableModels. Limits: takes no date and cannot reach analyses older than those 10, so it is not a monthly history; it returns no per-competitor scores. For competitor visibility use get_topic_visibility.
get_topic_mentions
ChatGPTBrand-mention leaderboard for a topic: every brand — YOUR brand AND each tracked competitor — ranked by how often LLMs name it when answering the topic prompts. This is the "share of voice by brand" / most-mentioned-brands ranking (use it to compare mention share across brands, e.g. GetMint vs Profound vs Peec AI vs Otterly). Returns mentions ([{ mention, count, percentage }], sorted by count DESC, where mention is the brand name and percentage is its 0-100 share of all mentions), mentionsDetailed (per-model { mention, modelId, count }), and pagination (total, page, limit, totalPages). Present each percentage with a "%". Limits: returns totals for one window, with no time series and no rolling average. Omitting startDate/endDate covers the last 6 months up to now, so for month-by-month figures call it once per month with that month as startDate/endDate. Never present figures from an unfiltered call as belonging to a specific month. Each share is cumulative over the window it was called with, for your brand and for competitors.
get_topic_shopping_overview
ChatGPTSHOPPING overview of one topic: how often the answers to its prompts came back with product carousels, how often a product of the brand is in them and at what position (its mean slot against the longest carousel), the competitor brands sharing the shelf, the merchants the offers land on and the top products, over a date window, with the topic name. Fed by the prompts of the topic whose modules include shopping, on ChatGPT and Alexa Shopping. This is the SHOPPING surface only: for the answer-engine brand-mention ranking of a topic use get_topic_mentions. Empty sections when no prompt of the topic feeds shopping or nothing ran in the window.
get_topic_sources
ChatGPTThe Core Sources and Other Sources tables of the Sources Explorer tab: every site (or URL) the AI cites on the topic, with its ownership (owned, competitor, external), citations, mentionRate (share of those citations whose answer names your brand), a sparkline over the topic's last 7 run days and lastCitedDay. group core = the sources the AI keeps citing run after run, cited in at least 80% of the last 10 runs of a prompt on a model (on average across the topic's prompts and models, or for at least two of them), whatever the date range; other = the rest. groups gives the size of both tables. Paged and sortable. Filters: startDate/endDate, models, tagIds (the on-screen ones by default), promptIds, search, ownership. Pass up to 5 tracked competitors to get their mention rate on each source.
get_topic_summary
ChatGPTQuick headline for a topic, YOUR brand only: summary.averageVisibilityScore and summary.averageCompetitionScore (0-100, averaged over the topic analyses of the last 30 days), report counts, lastAnalyzed, analysisRange, and trends (improving / declining / stable). Limits: the 30-day window is fixed and takes no date; it returns no competitor figures and no sentiment; the average is a single value, not a rolling series. For competitor scores use get_topic_visibility.
get_topic_visibility
ChatGPTVisibility report for a topic: how often LLMs mention YOUR brand AND each tracked competitor when answering the topic prompts. Returns visibilityScore (your brand, 0-100 percentage), scoreVariation (change vs the previous period, in percentage points; 0 = no change), competitors (each { name, visibilityScore }, ranked highest-first — use this to compare your brand against competitors like Profound, Peec AI, Otterly), modelBreakdown (per-model visibility), topMentions, totalPromptsTested, totalAnswers (answers the scores are computed from). Every score is a 0-100 percentage — present it with a "%". Limits: returns totals for one window, with no time series and no rolling average. Omitting startDate/endDate covers the last 6 months up to now, so for month-by-month figures call it once per month with that month as startDate/endDate. Never present figures from an unfiltered call as belonging to a specific month. dateRange ({ start, end }) gives the first and last analysis dates the scores cover — quote it so the user knows the period.
get_topic_visibility_by_model
ChatGPTVisibility of each tracked brand on each AI model of a topic, as the "Brand Visibility by Model" tab shows it: per brand its pooled level and margin, and byModel with each model's level and margin. A model has fewer answers than the pool, so its margin is wider: two models whose ranges overlap are not different. Filters: models, tagIds. Always give a score with its range (low–high). Brands in the same tier are too close to call: never call one ahead of another inside a tier. When derived is true, say the numbers are a combined estimate of the models asked for. Markets are not a filter of this read. When tooFewPrompts is true, lead with the warning that fewer than 10 prompts carry the selected tags, so the numbers are not significant, and draw no conclusion.
get_topic_visibility_prompts
ChatGPTYour visibility on each prompt of a topic, as the Prompt Visibility Ranking tab shows it, highest first: level (the latest 14-day reading) with its margin, runs (answers it is built on) and ratePct (share of those answers naming you). Use it to find the prompts where you are absent or strong. Filters: models, tagIds (the on-screen ones by default). One prompt has few answers, so its margin is wide: never compare two prompts whose ranges overlap. Always give a score with its range (low–high). Brands in the same tier are too close to call: never call one ahead of another inside a tier. When derived is true, say the numbers are a combined estimate of the models asked for. Markets are not a filter of this read. When tooFewPrompts is true, lead with the warning that fewer than 10 prompts carry the selected tags, so the numbers are not significant, and draw no conclusion.
list_briefs
ChatGPTList content briefs for a domain with pagination. Use this as the starting point to discover briefIds before calling other tools.
list_domain_reports
ChatGPTList the bi-weekly AI visibility reports of the organization, newest first: one report per domain every other Monday, covering the two weeks before. Each item carries the domain, the period and whether it is the newest report of its domain. Read one with get_domain_report.
list_domain_tags
ChatGPTThe prompt tags of a domain, each { id, name }. Use it to turn a tag the user names into the tagIds the visibility tools take.
list_domains
ChatGPTList every domain accessible to the authenticated organization. Typically the first call an agent makes — all other insights tools require a domainId.
list_explorer_citations
ChatGPTPaginated list of INDIVIDUAL citations — the exact source URLs LLMs cite, item by item (not just the aggregated domains from get_domain_explorer_aggregated). Each row carries link (the exact cited URL), linkDomain, the analysis type, and the model. Use the filters (the array filters take a list of strings) to drill from a domain to its exact pages: pass topicName: ["US - GEO - Transactionnel"] to scope to a topic (use the exact name from list_topics), linkDomain: ["reddit.com"] to list the exact Reddit threads, linkDomain: ["youtube.com"] for the exact videos, or linkDomain: ["g2.com"] to find the exact page to target for a mention. searchQuery substring-matches the PROMPT/QUERY text that produced the citation (not the cited URL — use linkDomain for that). HARD LIMIT: results are always scoped to the LAST 1 MONTH (the citation set is huge; a wider window would time out). To read further, pass the response's nextCursor back as cursor (unsorted calls only, same filters); page/pageSize also work.
list_topics
ChatGPTList every tracked topic attached to a domain.
trigger_content_generation
ChatGPTEnqueue article generation for a brief whose document has been finalized. Consumes CONTENT_GENERATION credits (scaled by word count). Returns immediately (202); poll with get_generation_status. Pass an idempotencyKey and reuse it when you retry a call that timed out or failed: the retry then returns the first call's result instead of generating and charging a second time. If the brief's metadata.brandKitId is set, generation applies the kit's tone, pronoun, style flags, author persona, and blocklist (blocked words/competitors are rewritten out during the polish step).