ProtoScrape gives developers a way to collect AI search observations and use them in their own products. Submit a query through an API, follow the job, and retrieve structured answer text and available citations. The workspace brings account access, API keys, jobs and credit usage into one place.
The first available engine is Google AI Mode, in early access. Google Search, ChatGPT and Perplexity are in validation and are not currently available for collection. This distinction matters when choosing what to build today.
For those research workflows, read the guides to ChatGPT search visibility, Google SERP data and Perplexity citation tracking. Each guide explains measurement choices and clearly states its engine’s current availability.
What comes back from the API
For a successful AI Mode job, the result includes the query, engine, collection timestamp, market metadata and a data object containing answer_text and citations. Each available citation includes a URL and its returned title. Some valid answers have no citations; an empty citation list is different from a failed job.
This format lets your application store observations without first interpreting a visual page. For example, a research tool can display the answer alongside its source links, while an SEO application can count which domains appear across a defined set of questions. The API supplies observations; your application defines the analysis.
A workflow your application can follow
Create an account, verify your email and generate a key in the workspace. Keep that key on your server. Use GET /v1/engines to check current availability, then submit a supported query to POST /v1/jobs with an idempotency key.
The response contains a job ID. Your application can check the job status and retrieve its result after success. A short wait option supports interactive requests without changing the underlying job workflow. The API documentation describes the endpoints, status codes and request limits.
Credits are reserved when a job is admitted and charged when it succeeds. Failed, cancelled or expired jobs release their reservation. Status checks and result retrieval do not create another collection charge.
Where the data is useful
- AI visibility research: observe whether a brand, product or topic appears in answers to a fixed question set.
- Citation analysis: study which pages and domains appear as sources, with a link back to each observation.
- Agency reporting: bring collected evidence into reports that explain the query, date, market and sample size.
- Developer products: add a server-side data source to your own dashboard, notebook or reporting pipeline.
These are workflows you can build with the API. ProtoScrape does not currently provide a finished rank tracker, a client report generator or an automatic content optimization service.
Current scope and practical limits
The current collection options are United States, English and desktop. Jobs and results are retained for approximately seven days, so export observations you want to keep for longer studies. Customer callback webhooks and batch submission are not available in this release.
An observation records one response at one time. It is not a guarantee of what every user sees, an official ranking score or evidence of the underlying model’s training sources. Repeat measurements and preserve the original evidence when drawing conclusions.
Start with one useful question
Choose a query that represents a real customer decision. Retrieve an answer, inspect its source links, and decide which fields your product needs to retain. Then expand to a small, repeatable sample. The Google AI Mode API guide walks through that first request, and the visibility measurement guide explains how to interpret a series of observations.