For drafting, research and reading documents, a general AI assistant is the right default: ChatGPT, Claude and Microsoft Copilot each score 5 of 5 on our five criteria, Gemini 4.5 and Perplexity 3.5. For a sourced property file, underwriting or a value opinion on one address, add Titleman (our product), which scores 4 and is used from inside the assistant over its connector.
Published by Titleman, which makes one of the tools compared. Updated 19 October 2026. Every general assistant is described from its vendor's own site or docs as read on 19 October 2026; we did not run them.
What commercial real estate analysis asks of an assistant
The work is a mix of reading and building: a rent roll and T-12 in, a sourced file on the property, a model out. A general assistant handles the reading and drafting. The open question is where the facts about a specific property come from, and whether you can trace each figure.
Scoring the six
Each tool is scored from its vendor's own pages: Yes counts 1, Partly 0.5, and no mention counts 0.
- A. Reads uploaded files and spreadsheets.
- B. Cites sources for what it returns.
- C. Business data controls stated: no training on business data, or admin and account controls.
- D. Connects to other tools through connectors or MCP.
- E. Builds or edits Excel or other spreadsheets.
| Tool | A | B | C | D | E | Score |
|---|---|---|---|---|---|---|
| ChatGPT | Yes | Yes | Yes | Yes | Yes | 5 |
| Claude | Yes | Yes | Yes | Yes | Yes | 5 |
| Microsoft Copilot | Yes | Yes | Yes | Yes | Yes | 5 |
| Gemini | Yes | Partly | Yes | Yes | Yes | 4.5 |
| Titleman | Yes | Yes | Partly | Partly | Yes | 4 |
| Perplexity | Partly | Yes | No | Yes | Yes | 3.5 |
Ties are ordered alphabetically, so the order among the top three is not a quality ranking. Titleman scores Partly on C because its docs state account-scoped sign-in and revocation but its connector docs state no SSO, and Partly on D because it connects to AI assistants and files in and out, with other integrations set up with our team. Perplexity scores No on C because the pages we could read do not state business data controls.
AI assistants compared
| Tool | Best for | What it does | Where it stops | Pricing model |
|---|---|---|---|---|
| ChatGPT | Drafting, research, spreadsheets | Reads XLSX, CSV and documents; deep research with sources; creates and edits spreadsheets | Not a source of property records; OpenAI says deep research can make incorrect inferences | Free and paid plans; see their site |
| Claude | Long documents and Excel edits | Reads Word, PDF, CSV and Excel; edits workbooks; web search with citations; connectors | Not a source of property records | Free and paid plans; see their site |
| Microsoft Copilot | Teams in Microsoft 365 | Excel formulas, charts and PivotTables; web answers; your organization's data | Not a source of property records | Licensing varies; see Microsoft's site |
| Gemini | Teams in Google Workspace | Reads uploaded files; Sheets; Drive and Gmail context; 100+ integrations in Gemini Enterprise | Citations not stated on pages read | Not stated on pages read |
| Titleman | Property file, underwriting, BOV | Sourced file and, on most runs, a live Excel model on an address, run from inside an assistant | Property only; no CRM, no lease-by-lease modeling | Not on this page |
| Perplexity | Web research with citations | Search with inline citations; Computer agent builds documents and spreadsheets | Data controls not stated on pages read | API billed at standard rates; see their site |
The ranked list
1. ChatGPT: best default for drafting and spreadsheets
OpenAI says ChatGPT reads spreadsheets (XLSX, XLS, CSV, TSV) and documents, and that ChatGPT Work, in Business and Enterprise plans, creates and edits docs, slide decks and spreadsheets. Deep research cites sources, and apps such as Google Drive and Slack can feed it. Business data is not trained on by default, with SAML SSO.
- Where it stops: it holds no records for your property, and custom MCP servers need a Business or Enterprise/Edu plan.
- Choose it instead if your team already works in ChatGPT.
2. Claude: best for long documents and workbook edits
Anthropic says Claude reads Word, PDF, CSV and Excel files, edits workbooks with tracked edits, searches the web with citations, and connects to apps including MCP servers. Admins set access by team and tool, and data is not used for training by default.
- Where it stops: custom connectors are not verified by Anthropic.
- Choose it instead if you work from long leases and reports.
3. Microsoft Copilot: best inside Microsoft 365
Microsoft describes Copilot in Excel with formulas, charts and PivotTables and web search with citations, plus your organization's data through Microsoft Graph, shown only to users who can already access it. Prompts and responses are not used to train its foundation models, and Copilot Studio agents can connect to MCP servers.
- Where it stops: you own the checking of any non-Microsoft tool you connect.
- Choose it instead if your firm lives in Excel and Teams.
4. Gemini: best inside Google Workspace
Google says Gemini takes uploaded documents and spreadsheets, up to 10 files per prompt, builds charts from a spreadsheet, and with Workspace fills tables in Sheets. Deep research can browse the web plus Gmail, Drive and Chat if you choose. Workspace company data is not used to train its models.
- Where it stops: citations are not stated on the pages we read.
- Choose it instead if your documents sit in Drive and Sheets.
5. Titleman: best for the property file (our product)
Titleman is not a general assistant. It is a property engine you open from Claude, ChatGPT or Microsoft Copilot Studio. It draws on 2,100+ sources and 337 workflows, takes Excel files up to 40 MB, runs a BOV at scale for up to 100 addresses per run, and usually returns Excel plus a PDF or deck, with every figure traced to a source or formula and gaps named. It gives no final price or buy or sell verdict. It underwrites the property, never the borrower.
- Where it stops: it scores below Gemini here because it states fewer data controls and integrations. The connector exposes six of eleven job types.
- Choose it instead if you need a sourced file on a specific address, which a general assistant does not hold.
6. Perplexity: best for web research with citations
Perplexity's docs describe search with inline citations, and its Computer agent browses, writes code and creates documents including spreadsheets, with 400+ connectors. We could not read perplexity.ai, so this score rests on docs.perplexity.ai.
- Where it stops: file upload and business data controls were not stated in the pages we could read.
- Choose it instead if the task is cited web research.
What these tools do not replace
An assistant, general or property-specific, does not replace a licensed appraisal, an engineer, an attorney or a title search. Titleman likewise does not replace a licensed appraisal, an engineer, an attorney, a title search, or lease-by-lease institutional modeling; its docs call its outputs drafts for a licensed professional.
A test to run
Take a deal you closed. Give each assistant the rent roll and ask for the owner of record, the last sale and any recorded debt. Note which answers cite a source, which say "not found," and which sound sure without one. Then see the MCP servers for real estate.
FAQ
What is the best AI assistant for commercial real estate analysis?
For drafting, research and reading documents, a general assistant is the right default. ChatGPT, Claude and Microsoft Copilot each score 5 of 5 on our criteria, and Gemini scores 4.5. For a sourced property file, underwriting or a value opinion on a specific address, add Titleman (our product), which scores 4 and is used from inside the assistant.
Can I just ask ChatGPT or Claude to underwrite my deal?
They can read a rent roll or T-12 you upload and draft a model, and both can build and edit spreadsheets. They do not hold a property's records for you, so you check each input yourself.
Which is better for commercial real estate, ChatGPT or Gemini?
On our five criteria ChatGPT scores 5 and Gemini 4.5; the one gap is that the Gemini pages we read do not state that answers carry citations. Pick the one that matches the tools your firm already uses, Google Workspace or not.
How do I use Titleman with Claude or ChatGPT?
Add Titleman's connector in Claude, ChatGPT or Microsoft Copilot Studio, sign in to Titleman and approve access. The assistant can then start and read Titleman runs on an address, and the reports usually come back with an Excel workbook and a PDF or deck.
If an AI assistant gives me a value, do I still need an appraiser?
Yes, where a licensed appraisal is required. Titleman returns a value opinion or BOV with each figure traced to a source or formula and does not replace a licensed appraisal. A general assistant's number is a draft to check, not an appraisal.
Sources
Read on 19 October 2026: ChatGPT overview, OpenAI business, file uploads, connectors, enterprise privacy, deep research, Claude overview, Claude custom connectors, Anthropic financial services, Gemini file uploads, Gemini deep research, Google Workspace AI, Gemini Enterprise, Microsoft 365 Copilot overview, Copilot privacy, Copilot in Excel, Copilot Studio MCP, Perplexity Computer docs, Titleman connector docs, titleman.ai. Perplexity's main site was not readable by our tools. Request a demo to see Titleman on one address.