aidentme / How it works

How aidentme works

aidentme is a platform that monitors brand presence in generative AI answers. It asks neural networks the same questions your buyers ask, saves the answers and turns them into measurable metrics and an action plan.

The data pipeline: from prompt to report

Setting up the study. In the dashboard you define the brand, competitors and a pool of scenario prompts. The task is queued and parallelised: 100 scenarios × 5 AI systems × repeated runs — hundreds of measurements per wave.
Two collection streams. Stream A — direct API calls to the models (ChatGPT, Claude, Gemini, GigaChat) with the version, parameters and date recorded. Stream B — emulation of live user sessions for AI assistants where the on-screen answer matters (Alice, Perplexity-style interfaces): headless browsers, regional proxy rotation, human-like pauses.
NLP parsing. Every saved answer goes through entity extraction (NER — your brand and competitors in all spellings), sentiment analysis, mention-order tracking and claim labeling.
Source mapping. Citations and quoted domains are extracted from the answers: the platform builds a map of the sites, review platforms and media the models assemble their opinion of your category from.
Aggregation and reporting. Metrics are computed per slice — model, market, language, scenario segment — and shown on the dashboard. Significant findings pass analyst review and receive an evidence status.

Raw answers are stored in full: any figure on the dashboard can be opened down to the exact text a user would have seen.

Three prompt types that mirror the buyer journey

An audit is useful when queries reproduce real choice situations — from problem awareness to purchase.

1. Problem prompt (top of funnel)

“How do I automate invoicing as a freelancer on a small budget?” — we check whether AI links your product to the user’s need. aidentme records whether the brand made it into the answer summary and computes Semantic Match — how close your content is to what the model cites.

2. Comparison prompt (middle of funnel)

“What is better for a mid-size agency — [competitor] or the alternatives?” — intercepting a competitor’s traffic. NLP isolates every mentioned brand and records Rank Position along with the model’s arguments for and against you.

3. Transactional prompt (purchase)

“Which CRM should a design studio buy with contract templates and a Telegram integration?” — hard functional criteria. The platform checks AI Recommendation Rate and parses the citations: which pages the model took the specs from.

Repeated runs

Every scenario runs several times: generative answers vary, and a metric without repetition is chance. Measurement waves are compared with each other, so trends and anomalies are visible.

What each team receives

Marketing and brand — AI Share of Voice against competitors and the brand association cloud.
PR and communications — the list of media and review platforms the models cite, plus negative claims with their sources.
SEO and content — ready recommendations: which pages and wording are needed for the models to start citing your materials.
Product and e-commerce — factual errors in specs, prices and distribution claims.

Get a platform demo

We will show the dashboard on your category: brands, competitors, prompts, metrics. ChatGPT / Claude / Gemini / Alice / GigaChat.

Request a demo