BLACK FRIDAY

50% off your first 2 months on every plan

Code RANKNOV26

Ends Monday 30 Nov, 23:59 CET

Claim the deal →

REST API · Updated November 2026

What can you do with theRankscale API for AI Search?

Chief of Staff, RankscaleFacts checked against the product and the OpenAPI specification, November 2026

The Rankscale REST API sends your AI search visibility data, down to every raw AI answer, as JSON into your reports, warehouse, alerts and agents, and sets up your tracking for you. As of November 2026 it is included on Pro+ (€229 a month), Growth (€385) and Enterprise (€780). It suits data teams, developers and agencies that report to clients. Our ChatGPT follow-up form study analysed 7,313 AI answers collected through this API.

How deep does it go? Down to every single AI answer, not just the scores built from them:

  • Every raw AI answer as JSON. The full answer text and every URL it cites, for every tracked run, with cursor pagination and date filters, so you can rebuild any number yourself.
  • 11 reporting endpoints, 9 kinds of data beyond a visibility score. Citations, AI source-box domains, query fan-out, sentiment with keyword clouds, shopping with buybox winners, ad placements with creatives, local and Maps results, ChatGPT follow-up forms and competitor metrics.
  • Citations sliced every way. By URL and by domain, per engine, per prompt, for your brand and for each competitor.
  • Full write access. Create and edit brands (including brand analysis for false positives and false negatives), topics, prompt research runs and up to 100 search terms per request; run tracking now; read your credit balance, runway and burn rate.
  • Built for production. 200 requests per minute per key, v1 only gets additive changes, and any breaking change is announced 6 months ahead with Deprecation and Sunset headers.
  • The same depth on 17+ AI engines, in every country and every language.

Want answers in Claude or ChatGPT instead of JSON? Use the Rankscale MCP server: no code and no API key, you ask in plain language, on every plan from Essentials (€17 a month, billed yearly).

On the free trial, ask our support team in the in-app chat to switch on the API.

API, MCP or Looker / Data Studio integrator: which do you need?

Use the MCP server when a person wants answers in a chat, the REST API when software needs the data on a schedule, and the Looker / Data Studio integrator when you want a client or leadership report without writing code. All three read the same data as your Rankscale dashboard.

MCP serverREST API (Metrics API)Looker / Data Studio integrator
Best forMarketers, account managers, leadership, anyone who works in an AI assistantData teams, agencies, developers, agents and automationsAgencies and in-house teams that report in Looker Studio
How you use itAsk in plain language in Claude, ChatGPT, Cursor, OpenAI Codex or any MCP clientJSON over HTTPS, 11 reporting endpoints plus setup endpointsCharts, tables and scorecards in Looker Studio, no code
Sign-inYour Rankscale login (OAuth 2.1 with PKCE). No API keyAPI key as a Bearer tokenThe same API key and your Brand ID
Can change setupYes, if you allow it. Never deletesYes: brands, search terms, topics, runsNo, read only
VolumeOne question at a time; the assistant pages through runs in batches200 requests per minute per keyUp to 10,000 records per request
PlansEvery plan from Essentials (€17 a month, billed yearly)From Pro+ (€229 a month): Pro+, Growth, EnterpriseFrom Pro (€99 a month)

REST API

What is the Rankscale Metrics API?

The Rankscale Metrics API is a REST API that returns your AI search visibility data as JSON and manages your tracking setup, at https://rankscale.ai/v1/metrics/. It is not a separate product: it reads the same data as your dashboard (it does not return Google keyword rankings: for classic search positions, use a SERP API) and manages brands, search terms, topics, prompt research and credits. It is included on Pro+ (€229 a month), Growth (€385) and Enterprise (€780), allows 200 requests per minute per API key, and every endpoint is described in the OpenAPI 3.1 specification.

reporting endpoints
11
search terms per create or update request. Bigger lists go in batches of 100: send the next batch when the last one returns, and a script or AI agent loads thousands this way
100
requests per minute per key
200
notice before any breaking change
6 months

How do you make your first API request?

In three steps: create an API key in Settings, Integrations; send it as a Bearer token; and post a JSON body with your brand ID to a reporting endpoint.

  1. In Rankscale, open Settings, then Integrations, and create an API key. A key has the same access as the person who created it, including their brand read and write permissions.
  2. Send the key in the header Authorization: Bearer <key>. If a tool cannot set headers, pass it as the api_key query parameter instead.
  3. Find your brand IDs with GET /v1/metrics/brands, then post a JSON body to any reporting endpoint.

Let your AI assistant write the requests. Open the Metrics API section of the help area in your dashboard, click Copy content and paste it into Claude, ChatGPT, Cursor or Claude Code. The assistant then knows every endpoint and parameter, and writes and runs the calls for the report you describe.

curl
# Dashboard metrics for the last 30 days, ChatGPT and Google AI Mode
curl -X POST "https://rankscale.ai/v1/metrics/report" \
  -H "Authorization: Bearer $RANKSCALE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "brandId": "brand_abc123",
    "timeFrame": "30d",
    "aggregation": "daily",
    "selectedEngine": ["chatgpt_gui", "google_ai_mode_gui"],
    "userTimezone": "Europe/Berlin"
  }'
curl
# Every raw run for two search terms in January, 100 at a time
curl -X POST "https://rankscale.ai/v1/metrics/executions" \
  -H "Authorization: Bearer $RANKSCALE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "searchTermIds": ["st_abc123", "st_def456"],
    "isoStartDate": "2026-01-01",
    "isoEndDate": "2026-01-31",
    "limit": 100
  }'

Request bodies based on the examples in the OpenAPI specification. Every response is JSON and carries the header X-API-Version: 1.

What does an API response look like?

Every response is JSON with "success": true and the result under data. A page of raw runs from /v1/metrics/executions lists each AI answer with its search term, engine, time, full response text, cited URLs, the companies it names and the visibility scores, plus the cursor for the next page.

JSON
{
  "success": true,
  "data": {
    "executions": [
      {
        "id": "exec_...",
        "searchTermId": "st_abc123",
        "term": "best ai visibility tool for agencies",
        "topicName": "Agencies",
        "tags": ["decision"],
        "engine": "chatgpt_gui",
        "executedAt": "2026-11-15T06:00:00Z",
        "response": "The full AI answer as text ...",
        "myBrand": "Your brand",
        "citations": ["https://...", "https://..."],
        "companies": [ { ... } ],
        "visibilityScores": { "latest": ..., "h24": ..., "d7": ..., "d30": ... }
      }
    ],
    "nextCursor": "...",
    "hasMore": true,
    "totalEstimate": ...
  }
}

Shape from the ExecutionsResponse schema in the OpenAPI specification; the values are placeholders. Pass nextCursor as cursor until hasMore is false. totalEstimate appears on the first page only. If a request contains a field the endpoint does not know, the response adds a warnings list, often with a "Did you mean" hint.

More per run, on request. In the Rankscale executions endpoint, includeMetrics adds the visibility scores of each run (latest, 24 hours, 7 days, 30 days), and includeRawProviderResponse adds the full payload the AI engine returned, for audits and your own parsing (OpenAPI specification).

Which data can you pull with the API?

There are 11 reporting endpoints, all POST. Each returns a default number of rows that you can raise; the citations endpoint returns up to 5,000 unique URLs, or every URL with uncapped=true.

EndpointWhat it returnsDefault size
/v1/metrics/reportDashboard metrics for a brand: history, plus metrics by topic, engine and competitorOne window
/v1/metrics/search-terms-reportA snapshot per search term, optionally with the answer textAll terms in scope
/v1/metrics/executionsRaw runs as JSON: full answer, cited URLs, companies mentioned. Cursor pagination and a date filterYour limit per page
/v1/metrics/citationsCited URLs and domains, by engine and by query, for your brand and competitors5,000 URLs
/v1/metrics/sourcesLeaderboard of domains shown in AI source boxes100 domains, 200 terms
/v1/metrics/sentimentSentiment for your brand and competitors, with keyword cloudsOne window
/v1/metrics/fanoutQuery fan-out: the searches engines run, and whether you are found200 queries, 50 domains
/v1/metrics/shoppingProducts, merchants and buybox winners in AI shopping results200 items, 100 merchants
/v1/metrics/adsAdvertiser placements, share of voice and creative samples100 advertisers
/v1/metrics/local-resultsLocal and Maps results: share, average position, detection rate and top places50 places, 100 prompts
/v1/metrics/follow-up-formsChatGPT follow-up form rate, top questions and pre-selected answers50 questions

Shopping, ads and follow-up forms come from the engines that show them: shopping from ChatGPT, Google AI Mode and Bing; ads from ChatGPT and Google AI Mode; follow-up forms from ChatGPT. The report endpoints use the same trailing averages as the charts on your brand dashboard, so the numbers match what your team sees in Rankscale.

Average position, top 3 and mentions, for you and each competitor. In the Rankscale API, /v1/metrics/report returns mentions and the top-3 rate for your brand, and for each competitor its visibility score, appearances, detection rate, citations, average sentiment, average rank, latest rank and top-3 rate (CompetitorMetrics in the OpenAPI specification). /v1/metrics/search-terms-report gives your average rank per search term. Sentiment comes per engine from /v1/metrics/sentiment.

Mentioned or cited? In the Rankscale API, every run from /v1/metrics/executions lists the brands the answer names (companies, myBrand) and, separately, the URLs it cites (citations). So you can tell an answer that only names you from one that also links to your site.

Share of voice. The Rankscale API returns share of voice for ad placements: ownShareOfVoice from /v1/metrics/ads. For organic AI answers it compares you with each competitor on the metrics above instead.

Ads in AI answers. The Rankscale API also reports the sponsored ads shown inside AI answers, from ChatGPT and Google AI Mode, the engines that show them. /v1/metrics/ads returns:

  • how many runs showed ads, out of the runs that could show them (totalRunsWithAds, totalAdEligibleRuns);
  • every advertiser with its domain, appearances, appearance rate, engines, search terms and sample creatives;
  • your own appearances, appearance rate and share of voice among the advertisers;
  • the runs with ads and the advertisers per day, for trend charts.

The Rankscale ads report returns the top 100 advertisers by default, or every advertiser with uncapped=true (AdsReportData in the OpenAPI specification).

Which AI engines and countries does the API cover?

Every engine you track, in every country and language, on every plan from Pro+ (€229 a month): the API returns whatever your workspace tracks, with no extra fee per engine and no limit on countries.

  • Engines. Fully flexible prompt tracking across ChatGPT, Google AI Mode, Copilot, Claude and 12 other AI Search Engines, every one on every plan. Filter any report by engine with selectedEngine; the full list of engine IDs is in the OpenAPI specification.
  • Countries and languages. No country or language limit on any plan. Each search term tracks one prompt on one engine in one country (an ISO country code) and language, so a report can compare the same question in Germany and the United States, or in German and English. If an engine does not support a country, the API says so with invalid_region before anything is created.
  • History. The Rankscale API reads your full history, as far back as your tracking goes. When an engine is retired, its runs stay readable, with the retirement date and the engine that replaced it. New search terms on a retired engine are refused with deprecated_engine.

The engines you can track, as listed in the Rankscale app. Interface engines capture what a person sees in the product; model API engines ask the model directly, and the ones marked "yes" can also run with web search.

EngineTypeWeb search option
ChatGPTInterfaceAs in the product
Google AI ModeInterfaceAs in the product
Google AI OverviewsInterfaceAs in the product
GeminiInterfaceAs in the product
Copilot (Bing AI Mode)InterfaceAs in the product
Perplexity Sonar, Sonar Pro, Sonar Reasoning ProModel APIBuilt in
OpenAI GPT-5.6 Terra, GPT-6.1 SolModel APIYes
Google Gemini 3.0F, 3.8F, 3.1PModel APIYes
Anthropic Claude 4.5H, 5.5S, 5.5HModel APIYes
Meta Muse SparkModel APINo
DeepSeek V3Model APINo
Mistral LargeModel APINo
Engine names as shown in the Rankscale engine picker, 16 November 2026. The engine IDs for selectedEngine are in the OpenAPI specification.

Does the API return what people see in ChatGPT, or answers from model APIs?

Both, and every run tells you which. Interface engines capture the answer a person sees in the product itself, including its sources, shopping and map results: chatgpt_gui, google_ai_mode_gui, google_gemini_gui and bing_ai_mode_gui, plus Google AI Overviews (google_ai_overview), taken from Google Search. Model API engines return what a model answers through its API: Perplexity Sonar, OpenAI GPT, Anthropic Claude, Google Gemini, DeepSeek, Mistral and Meta Muse Spark models.

Shopping, ads and ChatGPT follow-up forms come only from interface engines, because only the interface shows them. The engine field on every run from /v1/metrics/executions names the engine, so you can keep interface answers and model answers apart in your own reports.

How much tracking can the API handle, and what does a run cost?

Using the API is free: requests cost no credits, however much data you pull. Credits are spent only when search terms run, so your tracking volume is set by your plan's credits, and the 200 requests a minute per key only limit how fast you read and manage data.

  • Cost per run. A run typically costs 0.25 credits per engine; model API engines cost 1 to 4 credits per run, depending on the model and web search (changelog).
  • Worked example. Pro+ includes 3,000 credits a month. At 0.25 credits a run, that is 12,000 runs a month: for example 100 prompts, every day, on 4 engines (100 × 30 × 4 = 12,000).
  • Search terms. Unlimited on every plan, created up to 100 per request. Each one runs on its own schedule: hourly, every 2 hours, daily, every 2 days, weekly, every 2 weeks, monthly, every 2 months, or manually whenever you choose (interval and intervalEveryN in the OpenAPI specification).
  • More volume. Growth (5,500 credits) and Enterprise (12,000 credits) raise the budget; custom plans set their own. See the credit calculator.

How do you manage setup through the API?

Besides reading data, the API sets up and runs what you track, so onboarding, bulk changes and scheduled runs can be automated.

AreaEndpointsWhat you can do
Brands/v1/metrics/brandsList, create, edit and delete. Deleting asks you to repeat the brand ID, hides the brand and pauses its search terms
Brand analysis/brands/{id}/analysisFind false positives and false negatives, add or remove brand name variants, mark your own brand, recalculate metrics
Search terms/v1/metrics/search-termsList (up to 5,000, or all with uncapped=true), create or update up to 100 per request, delete, /activate, /deactivate, /run now with an idempotency key
Topics/v1/metrics/topicsList, create, edit and delete
Prompt research/v1/metrics/prompt-researchesStart a research run from your keywords, then fetch the suggested prompts when it completes
Credits/v1/metrics/creditsYour balance and credits in use; runway and burn rate with includeRunway=true

Running search terms through the API spends credits exactly as in the app. Credits per run depend on the engine: 0.25 for one prompt on one engine in the interface (GUI), more for API models (pricing).

Set up exactly what you track. When the Rankscale API creates search terms, one row with several engines in aiSearchEngines becomes one search term per engine; websearch switches web search on or off for the model API engines that offer it; and each term gets its country, language, topic, tags and schedule. Brands take up to 15 additional domains, plus a default country and language for new search terms (OpenAPI specification).

How do time frames, engines and pagination work?

Every reporting endpoint takes the same filters: a time frame (preset or custom dates), an engine, topics and tags. The executions endpoint pages through raw runs with a cursor.

  • Time frames. Use a preset timeFrame of 24h, 7d, 30d, 3m or 1y, or a custom window with isoStartDate and isoEndDate. The default is 7d.
  • Engines. Filter with selectedEngine, using engine IDs such as chatgpt_gui or google_ai_mode_gui. Engines that are no longer supported stay readable in your history, with their retirement date and replacement.
  • Topics, tags and prompts. Filter by topic, by tag or by a single prompt (selectedQuery). With several tags, tagMatchMode keeps search terms with any or all of them, and tagFilterMode keeps the ones that carry the tags or the ones that do not. __UNTAGGED__ selects search terms without tags, and _orphaned those without a topic.
  • Trailing average or latest run. The Rankscale report endpoints return the same trailing average as your dashboard charts by default; set showLastRunMetrics to true for the raw result of the latest run instead.
  • Earlier periods. Combine a preset timeFrame with periodOffset to read an earlier window of the same length, for example to compare this month with the month before.
  • Chart buckets. aggregation groups a time series by hour, day, week or month; it changes the chart buckets, not the averaging window.
  • Branded or non-branded prompts. The sources, fan-out, shopping, ads, local results and follow-up form reports take selectedBrandMode (all, branded or non-branded), so you can leave out the prompts that name your brand.
  • Pagination. The executions endpoint pages with a cursor; pass the cursor from each response to get the next page.
  • Unknown fields in a request are ignored and listed in a warnings array, so a typo never fails silently.

What are the rate limits, errors and versioning rules?

Each API key allows 200 requests per minute, every error comes back as JSON with a machine-readable code, and v1 only gets additive changes, with 6 months' notice before anything breaks.

Rate limits

200 requests per minute per API key. Every response carries X-RateLimit-Limit, X-RateLimit-Remaining and X-RateLimit-Reset; a 429 response adds Retry-After. Back off before the remaining count reaches zero.

Errors

Errors are always JSON, in the same shape, with a machine-readable code. Every response echoes an X-Request-Id; send it to support and we can find your request.

JSON
{
  "success": false,
  "error": { "code": "rate_limited", "message": "...", "details": { } }
}

Codes include auth_failed, validation_error, rate_limited, invalid_engine, deprecated_engine, invalid_region and conflict. Sentiment that is still being calculated returns 503; retry after 30 seconds.

Versioning

  • /v1 is the current and only version. New fields can be added to v1 at any time, so ignore fields you do not know.
  • Breaking changes ship as /v2, and v1 keeps working.
  • Anything removed or changed is announced in the changelog at least 6 months ahead. During that time responses carry Deprecation (RFC 9745) and Sunset (RFC 8594) headers.

Is the API ready for enterprise teams?

Yes: every request can be traced, and breaking changes come with 6 months' notice.

  • Traceability. Every response carries an X-Request-Id; errors come back as JSON with a machine-readable code.
  • Stability. v1 only gets additive changes. Breaking changes ship as v2, announced at least 6 months ahead, with Deprecation and Sunset headers.
  • Custom plans add a custom API, SSO and SLAs; talk to us.
  • Data protection. Rankscale is GDPR compliant. Rankscale GmbH is based in Vienna, Austria.

Yes. The Share Links API creates, edits, extends and revokes read-only dashboard links per brand, so clients and stakeholders see live data without an account.

  • Endpoints under /shareLinksAPI/: list brands, list and create links, edit or delete a link, and extend it by 30 days or reactivate it.
  • It uses its own key, separate from the Metrics API key. See Managing Share Links via the API.
  • Control what a client sees. Through the Rankscale Share Links API you choose the dashboard tabs a link shows (allowedTabs), its starting filters (time frame, engine, topic, prompt) and whether viewers may change them (lockFilters), and from Pro a saved custom dashboard view (dashboardViewId). Each link reports how often it was viewed.
  • Share links are on every plan. White-labeled links with your own logo are included from Growth (€385 a month) and on Enterprise (€780), where links can also be set to never expire.

Can you use it with the Looker / Data Studio integrator, Google Sheets, BigQuery or an agent?

Yes. The Looker / Data Studio integrator brings Rankscale data into Looker Studio (formerly Google Data Studio) without code. Google Sheets, BigQuery, Databricks, Power BI, Whatagraph, your own app or client portal, and any other tool read the data through the REST API, and AI agents connect through the MCP server.

Looker / Data Studio integrator

  • Dashboard Metrics: time series of your brand metrics, overall or broken down by topic or engine. Use it for scorecards and trends.
  • Search Term Executions: one row per AI answer, with the full response, brand detection, rank, sentiment, citations and competitors, up to 10,000 records per request. Use it for tables and drill-downs.
  • Set up without code: create the key in Settings, Integrations, copy a Rankscale report template, enter your Brand ID. You can schedule the report by email from Looker Studio. See the Looker / Data Studio integrator.
  • If Google shows "This app is blocked", your Google Workspace admin needs to allow the connector. Our Looker Studio FAQ has the steps.

Command line and agents

  • OpenClaw. An open-source skill runs GEO reports, engine profiles, content gap and citation analysis from your terminal, Slack, Telegram or Discord. Read the guide.
  • Claude Code and other coding agents. Connect the MCP server, or let the agent call the API with your key.

n8n, Make, Zapier and API clients

  • n8n, Make and Zapier. Rankscale has no native app for them; call the Rankscale API from their HTTP request step. Send POST to the endpoint, put your key in the Authorization: Bearer header and the filters in the JSON body, then route the result to Sheets, Slack or your CRM.
  • Postman, Insomnia and code generators. Import the Rankscale OpenAPI 3.1 specification and every endpoint, parameter and example appears as a ready request.

What do teams build with the API?

Here are 15 things teams build with the API, from client reports to spotting content gaps to warehouse exports. Each one comes with a prompt. Connect the API first (see your first request), then paste the prompt into Claude Code, Cursor or ChatGPT. Would you rather ask questions in a chat? See the MCP server use cases.

For agencies offering GEO and AI search services

Put AI visibility into the client reports you already send

GEO and AI search agencies · REST API or Looker / Data Studio integrator

Clients already get a monthly report from Looker Studio, Whatagraph or a Google Sheet. Agencies pull Rankscale data into that same report, so AI visibility sits next to SEO and paid results instead of in a separate login. Because you own the output, you can also write the labels in your client's language.

  1. Create an API key in Settings, Integrations.
  2. Pull /v1/metrics/report per client brand for the month, with breakdowns by topic, engine and competitor.
  3. Add /v1/metrics/citations to show which sources the engines used, and /v1/metrics/sentiment for how they describe the brand.
  4. Load it into your reporting tool, or use the Looker / Data Studio integrator and a Rankscale report template.
  • Using the Rankscale API, build the monthly AI visibility report for [client brand]. Pull /v1/metrics/report for the last 30 days and the 30 days before, broken down by engine and by topic, plus /v1/metrics/citations and /v1/metrics/sentiment for the same window. Give me a table with visibility, mentions, citations, average position and sentiment per engine, each with its change against the previous period. Add the 5 URLs cited most for the client and for its top 3 competitors. Then write three plain sentences for the client on what moved and why, and export everything as a CSV I can load into our reporting tool.

Uses/report/citations/sentimentor the Looker / Data Studio integrator

Set up a new client with hundreds of prompts and tags

GEO and AI search agencies onboarding large accounts · REST API

A new client often comes with 300 to 500 prompts, tagged by product, funnel stage, persona or country. Adding them by hand is slow and error-prone. The API creates them in batches, with their tags, topics and engines, and you check your credit runway before you switch tracking on.

  1. Create the topics with POST /v1/metrics/topics.
  2. Send up to 100 search terms per request to POST /v1/metrics/search-terms, with tags. One row for several engines becomes one search term per engine.
  3. Check GET /v1/metrics/credits?includeRunway=true to see how long your credits last at that volume.
  4. Activate the search terms with /activate, or run a sample now with /run to show the client first results the same day.
  • I am attaching a CSV with 400 prompts and a tag column. Using the Rankscale API, create them as search terms for brand [brand ID] on ChatGPT and Google AI Mode with POST /v1/metrics/search-terms, in batches of 100, and send each batch only after the previous one has returned. Keep every prompt's tags, and assign each search term to the topic that matches its tag, creating topics that do not exist yet. Do not activate or run anything. When you are done, give me the number created per topic and a list of any rows that failed, with the error code.

Uses/topics/search-terms/credits/activate/run

Check every client brand in one weekly run

GEO and AI search agencies with many client brands · REST API

Account leads want to know on Monday which clients need attention. A script loops over every brand in the workspace, pulls the week's metrics and flags the ones that dropped or sit below your threshold.

  1. List brands with GET /v1/metrics/brands.
  2. For each brand, pull /v1/metrics/report for 7d and compare with the week before.
  3. Post the flagged brands to Slack or email, with a link to each dashboard.
  • Using the Rankscale API, list every brand in my workspace with GET /v1/metrics/brands. For each brand, pull /v1/metrics/report for the last 7 days and the 7 days before. Flag every brand whose visibility score is below 50 or fell by more than 5 points. For each flagged brand, name the engine and the topic that dropped most, and the 3 search terms from /v1/metrics/search-terms-report that lost the most. Sort by the size of the drop and give me one line per brand that I can paste into Slack.

Uses/brands/reportexample question from the OpenClaw guide

GEO and AI search agencies · Share Links API

Clients should see their own brand without an account or access to other clients. The Share Links API creates, extends and revokes read-only dashboard links per brand, so a new client gets a link the day they sign, and expired links renew without anyone clicking through settings. White-labeled links are included from Growth (€385 a month) and on Enterprise (€780).

  • Using the Rankscale Share Links API (it has its own key), list my brands and the share links that exist for each. Create a read-only share link for every client brand that has none yet, and give me a table with client, link and expiry date. Then extend by 30 days every link that expires in the next 14 days, and list any link you could not create or extend, with the reason.

Uses/shareLinksAPI/share-links/extendsee share links

Show a prospect their AI visibility in a pitch

GEO and AI search agencies · REST API, manual runs

AI answers change from run to run, so one run is a snapshot, not proof. For a pitch, set the prospect's prompts to manual and run each one several times through the Rankscale API: the deck then shows how often the prospect is named, next to the competitors that win instead, without starting a schedule that keeps spending credits.

  1. Create the prospect's brand with /v1/metrics/brands and its prompts with /v1/metrics/search-terms, with the schedule set to manual.
  2. Run each search term several times with /run, one idempotency key per run.
  3. Pull every answer from /v1/metrics/executions and count who is named and cited.
  • Through the Rankscale API, create a brand for our prospect [BRAND, domain] and 20 search terms from the buying questions below, on ChatGPT, Google AI Mode and Copilot, with the schedule set to manual and the tag 'pitch-[client]'. Before running anything, show me the credit cost: 20 prompts x 3 engines x 5 runs x 0.25 credits = 75 credits. After I confirm, run each search term 5 times, then pull every run from /v1/metrics/executions. Build a one-page pitch summary: in how many of the 300 answers [BRAND] is named and cited, per engine; the 3 brands named most instead; the 5 prompts where [BRAND] never appears; and the 10 domains cited most for those prompts.

Uses/brands/search-terms/run/executions

For in-house marketing and SEO teams

Run a weekly AI visibility stand-up in Slack

In-house SEO and content teams · REST API with an agent

A scheduled job posts the week's overview to your team channel every Monday at 9:00, and the team asks follow-up questions in the thread. Our OpenClaw guide shows this flow with an open-source skill that calls four API endpoints in parallel.

  • Every Monday at 9:00, pull my Rankscale data for the last 7 days and the 7 days before through the API: /v1/metrics/report by engine and by topic, /v1/metrics/citations, /v1/metrics/sentiment and /v1/metrics/search-terms-report. Post a weekly overview to our #geo channel: overall visibility and its change; every engine and topic that dropped more than 3 points; the 5 search terms that lost the most, each with the brand that was named instead; and any domain that started citing a competitor this week. End with three content actions for the week, each tied to a named search term.
  • For the topic that dropped most this week, open the AI answers from /v1/metrics/executions that no longer name us. Which brands were named instead, which URLs were cited, and what do those pages say that ours does not?

Uses/report/citations/sentiment/search-terms

Get an alert when visibility drops

In-house SEO · REST API

A monthly report tells you about a drop four weeks late. A small script compares this week with last week per topic and engine, and posts to Slack when a topic falls by more than your threshold, with the search terms that moved.

  1. Pull /v1/metrics/report for 7d, with topic and engine breakdowns.
  2. Compare with the previous window and keep the topics that fell.
  3. For those topics, pull /v1/metrics/search-terms-report to name the prompts that lost visibility.
  • Write a script that runs every morning against the Rankscale API. It pulls /v1/metrics/report for the last 7 days and the 7 days before, broken down by topic and engine, and posts to our Slack webhook when a topic drops by more than 5 points on any engine. Each alert names the topic, the engine, the old and the new score, and the 5 search terms from /v1/metrics/search-terms-report that lost the most. Keep under 200 requests per minute, back off when X-RateLimit-Remaining runs low, and log the X-Request-Id of every failed call.

Uses/report/search-terms-report

See who advertises against you in ChatGPT and Google AI Mode

In-house marketing, paid search and e-commerce teams · REST API

Sponsored ads now appear inside AI answers, next to the brands the answer recommends. The Rankscale ads report shows which advertisers appear on your prompts, how often, on which engine and with which creative, so paid and organic teams see the same picture.

  • Pull /v1/metrics/ads from the Rankscale API for brand [BRAND ID] for the last 30 days. Show the share of eligible runs that carried ads per engine, the 10 advertisers with the highest appearance rate with their domains and sample creatives, and our own appearance rate and share of voice among advertisers. Then list the search terms where a competitor advertises but our brand is not named in the answer, using /v1/metrics/search-terms-report, and compare the ads per day with the 30 days before.

Uses/ads/search-terms-report

Put AI visibility next to your Search Console clicks

In-house SEO teams · REST API with Google Search Console

A page can lose clicks because an AI answer now answers the question, or gain them because engines started citing it. Join the URLs AI engines cite from Rankscale with the clicks and impressions per page from Google Search Console, and you see both in one table.

  1. Pull /v1/metrics/citations for your domain: every cited URL, by engine and by query.
  2. Pull clicks and impressions per page for the same dates from the Search Console API or a GA4 export.
  3. Join on the URL and compare the window with the one before.
  • Pull every URL on [our domain] that AI engines cited in the last 28 days from the Rankscale API (/v1/metrics/citations, by engine and query), and the clicks and impressions per page for the same 28 days from our Search Console export. Join them on the URL. Show one table: URL, times cited per engine, clicks, impressions, and the change of each against the 28 days before. Then list the pages that gained AI citations but lost clicks, and the pages that lost citations, with the queries behind each.

Uses/citations/search-terms-report

For content and SEO teams

Spot AI answer content gaps: the claims engines source from other websites

SEO and content teams · REST API with Claude, ChatGPT or Claude Code

An AI answer is a set of claims, and each claim links to the page it came from. When the same claim keeps linking to the same other website, that website owns a statement your page should own. Pull every AI answer for your tracked prompts with the executions endpoint, then let your AI assistant find those claims and rank them. The top of the list is what to add to your pages: the fact, the number or the comparison the engines are looking for.

  • Get all the AI answers for my brand from the last 30 days through the Rankscale API: page through /v1/metrics/executions until the cursor is empty. Split every answer into its claims and note the URL each claim is cited to. Find every part of those answers that links to a different website than mine, and keep the claims that are linked to another website on 2 or more engines. Give me a ranked overview of the claims that are consistently linked to another website: the claim, that website, how many answers and engines cite it for the claim, and whether any page on my site states it (yes, no or stated differently). For the top 10, tell me which of my pages should add the claim and under which question heading.

Uses/executions/citations

Turn your data into page changes with your own AI assistant

SEO teams and agencies · REST API with Claude

Teams pull the AI answers and their citations for prompts where they lose, then ask Claude to compare those answers with their own page and list what to change. Rankscale's page audits and recommendations do this inside the product; the API and MCP let you run the same analysis in your own workflow and at your own scale.

  1. Pull the runs for the prompts you lose with /v1/metrics/executions, including the full answer and cited URLs.
  2. Give Claude those answers and your page, and ask what the cited pages cover that yours does not.
  3. Track the same prompts after the change to see if you are cited more often.
  • For the search terms where my brand is not named, pull the last 10 runs per search term on ChatGPT and Google AI Mode with /v1/metrics/executions, including the full answer and the cited URLs. Read my page [URL] and the 3 pages those answers cite most. Tell me what the cited pages cover that mine does not: facts, numbers, comparisons, steps and the questions they answer. Give me the changes as a list: the section to add or rewrite, draft text for it, and the search terms it should win.

Uses/executions/citations

For data teams, e-commerce, multi-location brands and research

Load AI visibility into BigQuery or your warehouse

Data teams · REST API

Keep AI search next to SEO, paid and revenue data in one model. A daily job exports new runs with cursor pagination and a date filter, and a weekly job writes the aggregated metrics. Analysts can then join AI visibility to anything else you track.

  • Build a daily job that loads Rankscale data into BigQuery. Pull yesterday's runs from /v1/metrics/executions with the date filter, follow the cursor until it is empty, and append them to a raw table, deduplicated on each run's ID so a rerun never doubles rows. Once a week, write /v1/metrics/report and /v1/metrics/search-terms-report for the last 7 days into an aggregates table. Stay under 200 requests per minute, retry on 429 after the Retry-After header, and log every X-Request-Id.

Uses/executions(cursor paginated)/report/search-terms-report

Monitor products and locations in AI answers

E-commerce, retail and multi-location brands · REST API

Shopping data shows which products, merchants and prices AI engines show, and who wins the buybox. Ads data shows which advertisers appear in AI answers. Local results show how often a map pack appears, whether your locations are in it and which rivals take the spot.

  • Pull /v1/metrics/local-results for all my search terms for the last 30 days. List the search terms where an AI answer shows a map of local places without our business, with the places listed instead, their average position and the engine. Group the list by city where the search term names one, and tell me which of our locations are missing most often.
  • Pull /v1/metrics/shopping for this week and last week. For each of our products, show whether it appears in AI shopping results, at which merchants and prices, and who wins the buybox. List the competitor products that took our place since last week, and add the advertisers from /v1/metrics/ads that appear on the same search terms.

Uses/shopping/ads/local-results

Study the raw AI answers yourself

Analysts and researchers · REST API

The executions endpoint returns every tracked run as JSON, with the full AI answer, the cited URLs and the companies mentioned. Our ChatGPT follow-up form study (Rankscale Follow-up Forms data, October 2026) analysed 7,313 answers to 765 buying prompts collected this way; the method and sample are on the study page.

  • Export every run for the search terms tagged [tag] from the last 90 days with /v1/metrics/executions. For each answer, record the engine, the date, whether a ChatGPT follow-up form appeared, the brands named in the order they appear and the domains cited. Then give me, per engine: how often each brand is named, the share of answers with a follow-up form, and the 20 domains cited most, each with its run count so I can see which numbers rest on enough runs.

Uses/executions/follow-up-forms

Give your own agent Rankscale data

Teams building their own AI agents · REST API

Agents that write reports, plan content or brief account managers need live visibility data. Call the API from the agent's code, or connect the agent through the MCP server. That works for an agent on your laptop, such as Claude Code or OpenClaw, and for a hosted agent, such as Claude Managed Agents, which runs on Anthropic's servers on a schedule, so the report is ready without anyone's computer switched on. The OpenClaw guide shows a ready-made skill; our webinar "From Manual Work to Automations: Claude Code x Rankscale REST API & MCP" (55 minutes) walks through building one with Claude Code, and compares the MCP server and the API from 33:54.

  • Build an agent skill that answers questions about our AI visibility from the Rankscale API. It lists our brands, then pulls /v1/metrics/report, /v1/metrics/citations, /v1/metrics/sentiment and /v1/metrics/fanout for the time frame and engine the user asks about. It quotes the AI answer behind every number it reports, taken from /v1/metrics/executions, and it never activates or runs search terms without asking first.

Usesthe MCP server or any API endpoint

What should you check before you choose an AI visibility API?

Check where the answers come from, whether you get the full answer text or only scores, what a request costs and how far back the data goes. AI visibility APIs differ most on these points; here is each one, with the Rankscale answer.

Question to askWhat it changes for youRankscale REST API
Interface answers or model API answers?A model API can answer differently from the product people useBoth, and every run names its engine (see how)
Full answer text, or only scores?Without the answer you cannot check a numberFull answer, cited URLs and named brands for every run
Mentioned or cited?A name without a link sends no visitorsNamed brands and cited URLs are separate fields
Competitor metricsA score alone does not show who wins the answerVisibility score, detection rate, average rank, top 3 and citations per competitor; share of voice for ad placements
Engines, countries and languagesSome vendors charge extra per engine or countryAll tracked engines, all countries and languages, on every API plan
How often the data refreshesDaily data catches a drop four weeks earlier than a monthly reportPer search term: hourly, every 2 hours, daily, every 2 days, weekly, every 2 weeks, monthly, every 2 months, or manual runs
How far back you can readTrends need historyYour full history, as far back as your tracking goes, also on retired engines
Rate limitsA low limit slows warehouse loads200 requests per minute per key, with rate-limit headers
What a request costsSome tools charge per callReading is free; only runs use credits, from 0.25 per engine
Can it set up tracking?Write access lets a script onboard a clientYes: up to 100 search terms per request, topics, brands, run now
WebhooksPush or pull decides how you buildNo webhooks; schedule your own pulls after each run
Security and data protectionProcurement asks before anything elseGDPR compliant; SSO and SLAs on custom plans
StabilityA breaking change breaks your reportsv1 only gets additive changes; 6 months' notice before v2

Which plans include the API?

The REST API is included on Pro+ (€229 a month), Growth (€385) and Enterprise (€780). Every plan, from Essentials (€17 a month, billed yearly), includes the MCP server and tracks 17+ AI engines, in every country and every language.

  • Essentials€17/mobilled yearly only · $17/moMCP120 credits a month2 brand dashboards
  • Pro€99/mo€84 billed yearly · $99/mo, $84 yearlyMCPLooker / Data Studio integratorCSV and Sheets export1,200 credits a month10 brand dashboardsunlimited seats
  • REST APIPro+€229/mo€195 billed yearly · $229/mo, $195 yearlyMCPLooker / Data Studio integratorREST API3,000 credits a month25 brand dashboardsunlimited seats
  • Growth€385/mo€327 billed yearly · $385/mo, $327 yearlyMCPLooker / Data Studio integratorREST APIwhite-label links5,500 credits a month50 brand dashboardsunlimited seats
  • Enterprise€780/mo€663 billed yearly · $780/mo, $663 yearlyMCPLooker / Data Studio integratorREST APIwhite-label links12,000 credits a month100 brand dashboardsunlimited seats

As of November 2026 (Rankscale pricing, read 11 November 2026). Every plan includes every AI engine, every country and every language, and unlimited search terms; the free trial is 7 days on Pro. Custom plans add a custom API, SSO and SLAs; talk to us.

Each plan also includes a monthly credit budget for tracking. To see what your prompts, engines and schedule cost, use the credit calculator on the pricing page.

On the free trial, ask our support team in the in-app chat to switch on the API.

Frequently asked questions

Is the API included on Essentials or Pro?

No. The REST API starts on Pro+ at €229 a month (pricing). The MCP server is included on every plan from Essentials (€17 a month, billed yearly), and the Looker / Data Studio integrator from Pro (€99 a month).

Should I use the API or the MCP server?

Use the API when software needs the data on a schedule: client reports, a warehouse, alerts or bulk setup. Use the MCP server when a person wants answers in Claude, ChatGPT or Cursor without code. Both read the same data as your dashboard.

Does reading data through the API cost credits?

No. Credits are spent only when search terms run, whether you start them in the app or through the API.

Can I create hundreds of search terms at once?

Yes. The API creates or updates up to 100 search terms per request, with tags, topics, engines and schedule. For bigger lists, send the next batch of 100 when the last one returns. You can also import a CSV file in the app.

How do I check a visibility number against the AI answers behind it?

Pull the runs behind it. The executions endpoint returns every tracked run with the full AI answer, the URLs it cites and the companies it names, so you can recount any score yourself. A share of runs settles after about 10 to 15 runs per prompt; before that, one answer can move it by 10 to 15 points (Rankscale research across nine engines and seven topics) .

Is the data the same as in the dashboard?

Yes. The API, the MCP server and the Looker / Data Studio integrator read the same data as your dashboard. The executions endpoint returns the full AI answer and the URLs it cited, so you can check every number against the answer it came from. In Looker Studio, small differences can appear because Looker sums and averages differently.

How many requests can I make?

200 requests per minute per API key. The rate-limit headers on every response show how many you have left.

Can one API key read all my brands?

Yes. A key has the access of the person who created it, including their brand read and write permissions, so it reads every brand that person can see.

Will the API break my integration?

Not without notice. v1 only gets additive changes. Breaking changes ship as v2, announced at least 6 months ahead, with Deprecation and Sunset headers on every affected response.

Where is the full endpoint list?

In the OpenAPI specification, which also lists the MCP permission scopes. Setup guides are in the help center.

Is there real-time data or a webhook?

There are no webhooks. Tracking runs on the schedule you set, from hourly to monthly, so pull on the same rhythm: a scheduled job (cron, Airflow, a Make or n8n scenario, or an AI agent) calls the API after each run and picks up the new runs with the date filter and the cursor.

Can I get the API on a monthly plan?

Is there a cheaper way to get the data out than the API?

Yes. The MCP server lets an AI assistant read your data from Essentials (€17 a month, billed yearly), and the Looker / Data Studio integrator plus CSV and Sheets exports come with Pro (€99 a month). The API, from Pro+ (€229 a month), is for automated pulls, bulk setup and your own tools.

Is there an SDK?

There is no official SDK. Generate a client in your language from the OpenAPI 3.1 specification, or let your AI assistant write the calls: copy the Metrics API section from the help area in your dashboard into Claude, ChatGPT or Cursor.

Does the API return share of voice?

Only for ad placements: the Rankscale ads endpoint returns your share of voice. For organic AI answers, the Rankscale API compares you with each competitor on visibility score, detection rate, average rank, top-3 rate and citations, per engine and topic.

Is this an LLM observability API?

No. The Rankscale API tracks how AI search engines such as ChatGPT and Google AI Mode answer questions about brands: who is named, which sources are cited and how that changes. It does not monitor the tokens, latency or costs of your own AI app; that is what LLM observability tools do.

On the free trial, ask our support team in the in-app chat to switch on the API.


Changes: November 2026: rebuilt the API page. The MCP server has its own page at rankscale.ai/mcp.

Sources (read 16 November 2026): Rankscale OpenAPI specification, MCP docs, changelog, pricing, help center articles AI Assistants (MCP), Connecting Looker Studio and Looker Studio FAQ, and the OpenClaw guide.

See also: Looker / Data Studio integrator · MCP server · Pricing.