Chat Picker

AI Chat Tools for Urban Planners: Traffic Analysis and Neighborhood Design

Different file limits, accuracy cautions and policy safeguards shape this afternoon test of ChatGPT, Claude and Gemini for urban planning work.

Sources checked 2 Oct 2026

AI chat tools for urban planners can organize traffic data, synthesize public comments, compare zoning text and sketch scenarios, but the vendor pages we read—OpenAI’s data-analysis page, Claude’s upload page and Gemini’s file page—do not say that any assistant is validated for traffic simulation, zoning compliance or neighborhood design. Chat Picker has not tested ChatGPT, Claude or Gemini for this work, so use the afternoon plan below, not a ranking. Unsourced statistics and test claims from the earlier version have been removed.

Know these limits before you start

Treat every response as a draft. OpenAI’s accuracy note says ChatGPT can produce incorrect or misleading output, may sound confident and should not be a final source. Anthropic’s accuracy note says not to rely on Claude as the only source of truth and to inspect original sources behind web results. Gemini’s source guide says a Sources button appears when sources are available; Google’s safety guidelines describe LLMs as probabilistic and warn that Gemini may overgeneralize.

File limits decide what fits. As read on October 1, 2026, OpenAI’s File Upload FAQ lists a 512MB per-file ceiling, a 2M-token cap for each text or document file and three uploads per day for Free users. Claude’s upload page lists 500MB per chat file and 30MB per project file; PDFs up to 100 pages receive text and visual analysis, while PDFs from 101 to 1,000 pages receive text-only processing. Gemini’s file page lists up to 10 supported files per prompt, 100MB for non-video files and 2GB for video. Check the live pages before loading your packet.

The plan pages read on October 1, 2026 say ChatGPT offers an opt-out from training on chats on Free, Go, Plus and Pro. Claude offers one on Free, Pro and Max and says Team is not trained on by default. Gemini’s plans page does not state this setting, so verify the account control before use.

OpenAI’s usage policies prohibit unauthorized aggregation or distribution of private or sensitive information and prohibit automated high-stakes decisions in sensitive areas without human review. Anthropic’s Usage Policy classifies legal interpretation as high-risk and calls for qualified professional review. Google’s guidelines warn against factual inaccuracies that could cause significant real-world harm. Remove personal details from comment files and route zoning interpretations to an appropriate qualified professional.

The test plan

Run each step in a fresh chat with the same de-identified packet and plan. Save the originals, and keep one assistant’s answer out of the next assistant’s prompt.

  1. Traffic flow and congestion. Input: a turning-movement table, signal timings, a map image and field notes. Instruction: “Use only the attached files. Separate arithmetic findings from simulation assumptions, identify supported congestion locations, state units and missing fields, propose schematic signal changes, and cite the file and row behind each factual statement. Do not claim to have run a traffic simulation.” Look for traceable units, explicit assumptions and invented results. Log the plan, visible model label if shown, file versions, corrections and unsupported claims.

  2. Community engagement. Input: an anonymized public-comment CSV and a codebook. Instruction: “Group the attached comments by recurring theme. For each group, list the comment IDs included, summarize disagreements and identify comments that do not fit. Do not infer identity or sensitive traits.” Look for traceability and whether disagreements remain visible. Log omitted themes, unsupported characterizations and manual edits.

  3. Zoning compliance. Input: an ordinance excerpt, definitions and a requirements table. Instruction: “Use only the attached code. Quote the exact section for each requirement, distinguish direct text from inference, list ambiguities and draft questions for a planner. Do not provide legal advice.” Look for invented citations or unsupported legal conclusions. Log every citation checked, ambiguity raised and item sent for professional review.

  4. Scenario design. Input: a baseline map, design objectives, constraints and an accessibility brief. Instruction: “Propose several schematic alternatives. For each, list assumptions, trade-offs, constraints met, constraints missed and data needed for further study. Do not present an image or description as engineering approval.” Look for constraint coverage and reasons behind design choices. Log unsupported proposals and rework.

  5. Validation. Input: baseline metrics, observed conditions, model assumptions and the intended decision. Instruction: “Create a validation checklist for the proposed traffic or neighborhood change. Separate calibration data from holdout observations, identify what must be calculated outside the assistant and flag every missing source or assumption.” Look for a clear boundary between drafting and engineering validation. Log unresolved assumptions, required field checks and professional follow-up.

How to read your results

Use hard gates before preferences: privacy, traceable sources, correct units and a clear separation of calculation, simulation and design judgment. If a response fails a gate, do not advance that task with it.

Across responses that pass, compare your correction effort, file fit, plan limit and verification time. A polished answer is not validation. Make the decision task by task rather than naming an overall winner.

Where the plans differ

Chat Picker’s three-way comparison matrix, checked October 2, 2026, separates documented differences into Free plan, Low-cost tier, Main paid plan, Heavy-use plans, Team plan, Context window in the app, How usage limits are described, Training on your chats and Ads. Each matrix figure carries its vendor source and read date; unconfirmed cells are marked “Not verified.”

As read on October 1, 2026, OpenAI’s ChatGPT Plus page lists Plus at $20 monthly, Anthropic’s Claude pricing page lists Pro at $20 monthly or $17 monthly with annual billing and $200 up front, and Google’s AI plans page lists AI Pro at $19.99 monthly. Confirm current terms before subscribing.

In those same pages, OpenAI publishes separate instant and reasoning context limits by plan, Claude says context can reach 1M on every plan, and Gemini gives no app-token figure. The usage descriptions also differ: OpenAI and Claude provide qualitative limits without message counts, while Google describes compute-based limits and weekly caps.

Sources

For current prices and limits, see the dated comparison pages.

Compare assistants