compare_companies
ChatGPTCompare scorecards (EPS, sentiment %, visibility %, relevance) side by side for 2–10 individual profiles — markets, subsidiaries, or brands within the organization. Each profile is reported at ITS OWN latest measured period and labeled with it; the coverage note says when periods differ.
get_attribute_breakdown
ChatGPTEmployer brand attribute scorecard (Culture, Leadership, Compensation, Career Opportunities, Wellbeing & Balance, …) for the latest measured quarter(s): % of answers discussing each attribute, positive-sentiment %, share of all themes, change in points vs the previous measured period, and the top themes behind each. Use for deep-dive employer brand analysis. For a market-filtered view or example quotes use get_attribute_themes. Results carry _meta.periods (every measured quarter in the window) and lead with percentages; raw counts are under sample_size and are context only.
get_attribute_themes
ChatGPTUse for questions like 'what's our culture like in India?', 'how is our compensation perceived in Germany?', 'how does AI describe us for finance roles?' or 'why did wellbeing change?'. Filter by market (location) and/or job function (job_function); set by_job_function=true to see which functions hear an attribute most. Returns, per employer-brand attribute and per measured quarter: the % of answers discussing the attribute, its positive-sentiment %, its share of all themes, and the change in points vs the previous measured period. When a single attribute is requested it also returns real example themes with quote snippets and which AI platform said them, plus the sources cited in the answers that discuss that attribute (% of those answers citing each domain, with top_pages — the most-cited page URLs to link — association, not cause). Numbers match the PerceptionX dashboard. Attribute ids: mission-purpose-impact, compensation, company-culture, leadership, job-security, career-opportunities, wellbeing-balance, inclusion, innovation, application-communication, candidate-feedback, interview-experience, onboarding-experience (common aliases like 'pay', 'culture' or 'work-life balance' also resolve). Results carry _meta.periods (every measured quarter in the window) and lead with percentages; raw counts are under sample_size and are context only. Quote _meta.period_range and the matched market spellings when precision matters.
get_citations
ChatGPTWebsites/domains AI platforms cite when discussing the company, in the latest measured quarter(s): % of answers citing each domain, its share of all citations, the answer gap (% of answers citing it while the company was absent), and top_pages — the most-cited page URLs on that domain with their titles, for linking. Use domain_filter to drill into one source's pages (e.g. 'glassdoor.com' → which Glassdoor pages AI cites and how often). Use for 'how does Glassdoor appear in answers?', 'give me links to the pages AI cites about us'. Do NOT call get_responses alongside this for citation questions. For market-filtered or gap/opportunity ranking use get_sources. Results carry _meta.periods (every measured quarter in the window) and lead with percentages; raw counts are under sample_size and are context only.
get_company_metrics
ChatGPTJust the scorecard for the latest measured quarter: EPS, positive-sentiment %, visibility %, relevance, change in points vs the previous measured period, and sample sizes. EPS = 50% sentiment + 30% visibility + 20% relevance. Numbers match the dashboard (brand scope by default).
get_company_overview
ChatGPTThe dashboard's default view in one call: for the LATEST measured quarter of the brand scope — EPS, positive-sentiment %, visibility %, relevance, and change in points vs the previous measured period — plus the top attributes by % of answers discussing them, top themes, top competitors and top sources (each as % of answers). Use this as your default first tool when a user asks how a company is doing. For history use get_trends; for a market use the market-aware tools. Results carry _meta.periods (every measured quarter in the window) and lead with percentages; raw counts are under sample_size and are context only.
get_competitor_landscape
ChatGPTCompetitors named by AI platforms across the brand scope, filtered by market and/or job function: % of answers naming each competitor and % naming them alongside the company, with per-prompt-type context. Optionally pass attribute_id (e.g. 'compensation' or the alias 'pay') to get the attribute lens: share-of-voice on prompts about that attribute ('who gets named when pay comes up') plus early competitor-sentiment themes where available. IMPORTANT: share-of-voice is who gets NAMED, not who is rated better — say so when answering 'who is our top competitor for X'. Names are canonicalized; job boards and the company itself are excluded from competitor lists. Results carry _meta.periods (every measured quarter in the window) and lead with percentages; raw counts are under sample_size and are context only.
get_competitors
ChatGPTCompetitors AI platforms bring up when answering about the company, in the latest measured quarter(s): % of answers naming each competitor and % naming them alongside the company. Brand scope by default. For a market or attribute lens ('who is our top competitor for pay?') use get_competitor_landscape. Results carry _meta.periods (every measured quarter in the window) and lead with percentages; raw counts are under sample_size and are context only.
get_model_breakdown
ChatGPTHow each tracked AI platform (ChatGPT, Perplexity, Google AI Overviews, Google AI Mode, and Claude from September 2026) perceives the company in the latest measured quarter(s): visibility % and positive-sentiment % per platform. Useful for 'which platform mentions us least / is least favorable'. Results carry _meta.periods (every measured quarter in the window) and lead with percentages; raw counts are under sample_size and are context only.
get_negative_drivers
ChatGPTWhere this organization's negative opinions come from, for the latest measured quarter with the previous one for comparison: for each employer brand attribute, its share of all negative opinions (negative themes on the attribute as a % of every negative theme in scope), its positive-sentiment %, and the share of its negatives that sit in answers citing one of the organization's own review-profile pages (Glassdoor, Indeed, kununu, AmbitionBox, Seek, Jobstreet, Comparably and the smaller review sites), plus the overall share of negatives in such answers, the four candidate experience attributes combined, and the review-profile pages AI cites most (link these). Filter by market (location) and/or job function. Use for 'where do our negatives come from', 'which topics hold most of the negative opinion', 'how much of the negative is review-site driven', and as the first step before recommending what to fix (then get_attribute_themes for the topic, get_report_actions and get_playbook). Change is in percentage points vs the previous measured period. The review-profile shares are computed on a random sample of the answers that carry negative opinions, and sample_size says how many. Results carry _meta.periods (every measured quarter in the window) and lead with percentages; raw counts are under sample_size and are context only.
get_playbook
ChatGPTPerceptionX's employer reputation playbook: how to move the score, how to diagnose a weak topic, who at an employer can change what, what a strong career site and job posting contain and where a block goes on a posting so AI and job boards read it, AI readability, pay transparency law, candidate experience, employer lists and awards, review platform capabilities and rules, and how to write a recommendation. Use for any 'what should we do', 'how do we fix', 'how does Glassdoor / kununu / our career site work' question, together with the data tools. This is guidance, not data about the organization: never present it as a measurement, and take every figure from the data tools. Dated facts (laws, platform features, list deadlines) carry an as-of date: say it and tell the user to check before acting.
get_report_actions
ChatGPTWhat PerceptionX reported and recommended to this organization: the latest published report snapshot (period, as-of date, scope, headline figures, market / attribute / source breakdowns, and which metrics are final or provisional), the numbered actions with problem, what AI says, where AI reads it (source and coverage), exact URLs, what to do, lane, timing, what it lifts, verified facts with check dates, owner and status, plus watch items. Use for 'what should we fix first', 'what did you recommend', 'why is X low', 'what is the status of action 2', 'is our score final'. Cite the action number, the report period and the as-of date (cite_as). Say 'provisional' when quoting a metric whose metric_status is provisional. Pair with the live tools when the user wants current data, and give both figures with their dates if they differ. Read-only: status and owner are set by PerceptionX. No company_id needed.
get_responses
ChatGPTGet actual AI answer texts for one company profile, newest first. Essential for qualitative questions like 'what do AI platforms say about X', 'how is the culture described', 'what are the negatives'. Returns the answer text, which AI platform wrote it, sentiment, the prompt asked and the period. Filter by prompt_type for focused analysis.
get_sources
ChatGPTWhich websites AI platforms cite when answering about the company, filtered by market and/or job function — led by the % of answers citing each domain, with the ANSWER GAP: % of answers that cited the domain while the company was NOT mentioned. Set gap_only=true to rank by that gap — the outreach-opportunity list ('sources answering candidate questions in your space without you'). Domains are canonicalized (glassdoor.de/.ie/.com collapse to one). Every source carries top_pages — its most-cited page URLs with titles — so answers can link the exact pages. Use for 'which sources matter in Germany?', 'where should we be mentioned but aren't?', 'link me the pages AI cites'. Results carry _meta.periods (every measured quarter in the window) and lead with percentages; raw counts are under sample_size and are context only.
get_themes
ChatGPTRecurring themes extracted from AI answers in the latest measured quarter(s): each theme with the % of answers mentioning it, its positive-sentiment %, the attributes it belongs to and which AI platforms raised it — plus an attribute summary (% of answers discussing each attribute, sentiment %, change vs the previous measured period). For a market-specific view prefer get_attribute_themes. Results carry _meta.periods (every measured quarter in the window) and lead with percentages; raw counts are under sample_size and are context only.
get_trends
ChatGPTSeries by measured quarter for one metric — 'visibility' (% of answers mentioning the company), 'sentiment' (positive-sentiment %), or 'citations' (citations per answer) — optionally filtered by market and/or job function. Returns the series plus change in points vs the previous measured period and since the first. Every listed period is a complete collection wave; a period is marked '(in progress)' only while its wave is still being collected. Use for 'is our sentiment in Germany improving?', 'visibility trend this year'. Results carry _meta.periods (every measured quarter in the window) and lead with percentages; raw counts are under sample_size and are context only.
get_visibility
ChatGPTVisibility = % of AI answers that mention the company by name. Filter by market and/or job function and get the series by measured quarter with the change in points vs the previous measured period; set by_model=true to split by AI platform (ChatGPT, Perplexity, Google AI Overviews, Google AI Mode, and Claude from September 2026) or by_job_function=true to split by job function. Use for 'how visible are we in Japan?', 'is our visibility improving?', 'which AI platform mentions us least?', 'which job functions see us least?'. Numbers match the dashboard rollups. Results carry _meta.periods (every measured quarter in the window) and lead with percentages; raw counts are under sample_size and are context only.
list_companies
ChatGPTList all companies/market profiles in the user's organization. Always call this first if you don't already know the company IDs. Returns id, name, country, industries, latest_period (the profile's most recent measured quarter), measured_periods (how many quarters have data) and total_responses (a size cue for choosing a profile). Same-name profiles are one brand measured per market; the other tools aggregate them like the dashboard.
search_responses
ChatGPTFull-text search through AI answer texts for one company profile to find answers that mention specific topics, keywords, or themes (newest first). Use this when the user asks about a very specific topic.
