The one public review of WRDS's Ask AI chatbot finds a lookup tool with a six-question limit

Almost nobody has reviewed the official WRDS chatbot in public. A search for reviews and forum threads turns up one substantial assessment: a June 2026 review by Lim Hwee Ming, a senior librarian at Singapore Management University, covering AI features in WRDS, Bloomberg and Fitch Connect.

A first test returned the wrong documents

	What the reviewers report
What it is	The WRDS Ask AI Chatbot, beta version 0.5, opened from the top right of the WRDS site
First test	A query on “Capitalized Software” returned datasets outside the school’s subscription and “results with wrong indexed documents”
Limits	Up to six questions a session, with prompts under 200 characters recommended
Complex questions	Starting with one can produce “vague and repetitive responses”
Reasoning	Visible in a details panel, but it “can sometimes lack precision, with ambiguous interpretations”
Best at	Identifying data variables, understanding definitions and comparing datasets
Their advice	Verify every output “against official WRDS documentation”

The reviewers describe a tool that improves with specific prompts and follow-up questions, and they place it with Bloomberg’s and Fitch’s features as an “assistive layer” for exploration.

The chatbot does not pull data or write code

Nothing in the review describes the chatbot pulling data, running a query or writing analysis code. It tells you where a variable lives, and you fetch the data the usual way.

Researchers are building the missing part themselves. The open source page lists eight public projects, all created this year, that connect WRDS to coding agents: skills that hold the schemas and identifiers, and MCP servers that expose CRSP, Compustat and IBES as tools.

WRDS prohibits loading its data into LLMs

WRDS publishes a policy on generative AI. It says that “loading Data retrieved from WRDS into LLMs and other Generative A.I. tools is prohibited,” with exceptions from named vendors for protected, enterprise versions of AI tools. The policy covers the community projects too. The licensed data guide explains how to stay inside it.

The evidence is one review of a beta

The review covers beta version 0.5, is four months old and comes from a single library.