Web research & Quanty Connect
AI columns can read the live internet — not just the row's documents. This guide covers how Quanty decides how to search, how a sub-sheet builds a whole researched dataset, and how Quanty Connect adds premium data sources to a run.
For the basics of the Data source selector (documents, internet, or both), see Columns & AI extraction.
The search method: Auto by default
Every web-source column has a Search method setting. The default, Auto — AI picks, lets Quanty choose the best way to fetch data for each row:
- If the task names specific pages (a URL in the prompt or in the row), Quanty reads those pages directly instead of searching for them.
- A simple current fact — a price, a rate, a name — gets a fast search.
- A typical lookup gets a standard search.
- A question that needs several sources compared gets a thorough search.
- A hard, multi-step question gets deep research: an agent plans its own searches, reads many sources, and returns one verified, grounded answer.
You can also pin the method yourself: Fast, Thorough, or Deep research. The method that was actually used — and why — is recorded with every value in the cell details panel.
Deep research reads many sources and reasons across them, so it takes minutes rather than seconds per row, and costs more. Every run has a hard spending cap.
Researched datasets in sub-sheets
A Sub-sheet column with the internet as its data source builds a whole dataset per row — one verified item per sub-row.
Give it a research task as the prompt:
Find 25+ US AI infrastructure companies that have raised $5M+ and are hiring designers. For each, include its infrastructure category, latest funding, the design hiring signal, and a source URL.
and sub-columns for the fields you want (company, category, funding, hiring signal, source) — or let AI propose the schema from the prompt. When the row runs, deep research fills the sub-sheet: every item verified against sources, duplicates collapsed, missing fields marked honestly as not found.
You can also ask for this in Chat: describe the dataset you want and the assistant builds the sub-sheet column, adds a row for the topic, and runs the research — or lays out a whole page of web columns for a broader brief.
Quanty Connect: premium data sources
Some answers don't live on public web pages — verified company registries, web traffic estimates, financial fundamentals. Quanty Connect attaches premium data sources to a research run: the research agent queries those databases alongside the live web and fuses everything into one grounded answer.
Enable sources per column in the Quanty Connect section of the web settings. Selecting any source switches the column to deep research, since partner data is only available there.
| Source | What it offers |
|---|---|
| Fiber.ai | B2B database of companies and people — lead discovery and contact research |
| Similarweb | Web analytics — traffic estimates, global rankings, competitor discovery |
| Baselayer | Compliance & KYB — verify US businesses: officers, registrations, risk signals |
| Affiliate.com | Commerce — product catalog search with pricing, brands and merchant links |
| Particle | Media intelligence — podcast transcripts with speaker attribution and timestamps |
| Financial Datasets | Finance — prices, fundamentals, earnings, SEC filings and ownership for 27,000+ US tickers |
| Jinko | Travel — destination discovery ranked by lowest available fare |
Available on request: Crunchbase, Definitive Healthcare, Faraday AI, Harmonic, Intellizence, Kernel and ZoomInfo. Contact us to enable them for your workspace.
A typical Connect column: "Estimated monthly web traffic of @Company" with Similarweb attached — the agent pulls the analytics figure from the partner database instead of guessing from a web page, and cites where it came from.
Each premium source call adds a small per-call fee on top of the research run, billed from your plan's tokens like every other web search. The full cost of a run appears on the usage page under Web search.
Cost and provenance
- Every web search, page read, and research run is billed from your plan's tokens and listed separately on the usage page.
- Every value keeps its provenance: the query, the method used, the pages read (title, address, publication date) and — for research runs — the sources behind each field, all visible in the cell details panel and the audit trail.