Company

We built the instrument
we wanted to use.

ResourceAI audits the public signals that search and AI systems can retrieve about a brand. It combines a rendered-page crawl, programmatic measurements, enabled provider checks, and a report whose claims remain traceable to their evidence.

The problem we saw

AI assistants have become another discovery surface alongside conventional search. Their answers can mention, omit, confuse, or misdescribe a brand, while traditional ranking tools do not expose every retrieval and generation mechanism involved.

A useful audit therefore has to keep different kinds of evidence separate: what the page contains, what structured sources resolve, which optional providers responded, and what could not be measured.

ResourceAI was built to make that evidence inspectable.

What we actually do

ResourceAI validates one submitted public URL, renders the page, performs HTTP and structured-data checks, classifies the brand when the classifier is available, and evaluates 22 declared score dimensions. Fifteen capability families can be independently enabled or disabled.

The output is a browser report with an editorial summary, dimension scores, an issue list, and a Receipts view. Each capability records whether its result was real, degraded, stubbed, disabled, or errored. The report can be printed or saved as a PDF.

Duration and measurement coverage vary with queue depth, page response time, audit scale, enabled capabilities, source fallbacks, and provider availability. The report records the resulting limits instead of hiding them.

How we're different

Mechanistic, not vibes

The crawl, structured data, tokenization, retrieval, source, and provider evidence remain separate so one generated answer does not stand in for the whole audit.

Provider receipts

Adapters exist for OpenAI, Anthropic, Gemini, Perplexity, and DeepSeek. The report names providers only when their calls actually ran and records usage and failures.

Explicit capability quality

The 15 capability families are feature-gated. A disabled, degraded, or stub result is labelled as such and must not be interpreted as a confident measurement.

Fix queue, not just findings

Every row in the dossier pairs a problem with a fix, an owner, a severity, and an estimated time to ship. Your team knows what to do Monday morning.

The team

The implementation is built with Next.js, BullMQ, Redis, Supabase, Playwright, pgvector, and provider-specific adapters. The repository contains the web application, worker, migrations, source waterfall, measurement modules, and verification tests.

Fast Scans are free and Deep Scans cost ₹49 (introductory) so the product proves itself the same way its reports do: with receipts. Nothing about the method changes with price — a paid run and a free run obey the same evidence rules.

If that question matters to you, we should talk.

Contact

General inquiries: hello@resourceai.in

Press: press@resourceai.in

Teams & agencies: enterprise@resourceai.in