Semantic Search is now live on web, bringing the platform onto the same search model that powers the Grata MCP and agentic workflows:
Free-text search that maps a description straight to company results, without converting it to keywords first
Free text and filters on the same page – users can move between describing what they want and narrowing with filters without starting over
New Target Criteria filter that uses semantic search to define what a target company does
Terms filter now defaults to Mentioned, since Target Criteria handles defining the core business best
Saved searches and signals supported
Search links will carry over between the platform and MCP, since both now run on the same model
Chat has moved off the search page to the new Grata Intelligence homepage
Accessible on the Companies page on web for all accounts on across all plans
Purpose & Use Cases
Target list building: As a dealmaker, I need to describe the kind of company I’m looking for in plain language and get a complete, relevant list back so I can start outreach faster
Market sizing and mapping: As a dealmaker scoping a space, I need to see the full universe of companies in it, including smaller private businesses deep in the results, so I can trust my market map and sizing
Building an investment thesis: As an investor exploring a new sector, I need to quickly surface the key players and adjacent niches so I can pressure-test a thesis before committing time to it
Finding comparables: As a banker or corp dev lead, I need to find companies that do the same thing as a given business so I can build comp sets and identify likely buyers or targets
Benefits
Faster target building – less time sifting, more time spotting real targets
Better market sizing – results are closer to the real universe, so users can trust our search is using our private company data to its full potential
Consistent results everywhere – the same search returns the same companies in the platform, MCP, or an agentic workflow, so work started in one carries into the other
Easier to use – users describe the companies they want the way they’d describe them to a colleague, with no keyword or filter translation needed
Results & Improvements
~2.4x as many relevant companies across the full result set (143% improvement), mostly from later pages – much less noise deep in the list
Relevance in the first 50 results up another 21% since semantic search launch via the MCP, from model updates made since
Key Definitions
Semantic Search: Search that interprets the meaning of a free-text description and returns companies that match it, without relying on exact keyword matches
Target Criteria: Filter that uses semantic search to define the core business of the companies a user is looking for
Terms (Mentioned): Filter that matches companies whose profiles mention specific words or phrases. Now defaults to Mentioned, for supporting terms alongside Target Criteria
FAQs
Who has access?
All accounts
What are best practices for using semantic search?
Describe what the companies do in as much detail as you’d like – products, services, end markets, business models, or niche specialties. The more specific the description, the more precise the results
Use filters if you want maximum control over the result set for firmographic criteria like location, employee count, revenue, and ownership – you can still write these out with free text, but for more precision you can always use the existing filters
Combine the two: describe the space in free text, then narrow with filters on the same page
Refine as you go – add or adjust details in the description to sharpen results, and e.g. use Terms (Mentioned) if you want to narrow down to companies that say specific words or phrases directly on their website
Example: “vertical SaaS for veterinary clinics, including practice management and client communication software,” then filter to US-based, 20–200 employees, and privately held
Is this the same search as the MCP?
Yes – one model now powers the platform, MCP, and agentic workflows
Where did Grata AI Chat go?
Chat is no longer on the search page. Users can chat through the new Grata Intelligence home page
Where did the keyword and filter translation go?
Previously, chat translated a free-text query into keywords and filters and displayed them. Semantic search maps free text straight to company results, so there’s no translation step to show
Users can still adjust their free text query and/or add filters alongside their free-text search on the same page to narrow results
Results continue to show evidence in tile view and the Evidence tab when clicking on the profile, so the user knows exactly why results match their query
Can users access their previous agentic chat history?
Previous chat history wasn’t carried over and won’t be accessible in the platform
If you need to access a specific past session, reach out to support@grata.com for assistance
Why does Terms default to Mentioned now?
The new Target Criteria filter does a better job of defining the core business with semantic search, so Terms works best as a supporting filter
Will existing saved searches still work?
Yes, this is backward compatible with all past searches
How was the search quality improvement measured?
We ran a set of representative customer searches through both the current agentic search and the new model, and scored the relevance of companies returned on the first, middle, and last pages of results