Three independent AI reviews — ChatGPT, Gemini, and Claude — scored gunnisoncounty.org's assessor pages against the same 100‑point civic-service rubric between September 15 and September 28, 2026. Total scores land between 60 and 78. All three flag the same underlying problem from different angles: the site is strong on data tools, weak on keeping deadlines and forms current.
Each bar is that category's score as a share of its rubric maximum, so categories worth different points compare fairly. Claude did not test usability or accessibility directly and marked both as estimates from the assessor office's own internal review — shown hatched.
Only the Gemini review scored enough counties to support a real ranking — 24 of Colorado's 64, chosen to span Front Range, mountain-resort, and rural jurisdictions. The other 40 counties were catalogued but not scored, and are excluded here rather than shown as zeros.
The ten recommendations below merge overlapping findings from all three reviews into one prioritized list, so duplicate advice ("fix the deadline," said three ways) becomes one line item.
Every review independently called out the same strengths — the fixes above are about currency and completion, not a rebuild.
Review the public-facing assessor websites for all 64 Colorado counties. My main goal is to understand where Gunnison County stands statewide and how its website could better serve property owners.
| Category | Pts | Evaluate |
|---|---|---|
| Findability and organization | 15 | How easily can a resident find property search, appeals, exemptions, forms, and contact information? Are navigation labels clear? |
| Property search and public data | 20 | Can a user find a parcel by address or account number and view useful values, property details, sales information, and maps? Are the data and their dates explained? |
| Self-service tasks | 20 | Can an owner find and complete or submit common requests, including an appeal, address change, exemption application, property-record correction, or information request? |
| Explanations and transparency | 15 | Does the site explain how property is valued, why a value may change, what the assessor controls, and how value relates to taxes? |
| Usability across devices | 10 | Do menus, searches, forms, and linked portals work clearly on desktop and mobile? |
| Accessibility | 10 | Are core tasks usable with a keyboard, readable text and contrast, meaningful link labels, and accessible forms or documents? |
| Currency and trust | 10 | Are deadlines, rates, notices, contact details, and resources current and clearly dated where appropriate? |
Assign letter grades of A (90–100), B (80–89), C (70–79), D (60–69), or F (below 60). Do not award an A if property search is unusable, appeal instructions are materially wrong or missing, or a major accessibility barrier blocks a core task. State the reason if this rule limits a county's grade.
Give every county a numeric score, letter grade, category scores, and brief evidence with links to the pages or tools examined. Distinguish missing, broken, and could not verify. Do not assume a search tool works merely because its link exists. Note when a task moves to a treasurer, GIS, vendor, or other external portal, and assess whether that handoff is understandable. Do not count a service as missing if it is available through a clearly linked external portal.
For each category, explain what earned or lost points using observed evidence. Apply the same scoring standard to every county: award full points when the relevant information or task is easy to find, clear, current, and usable; partial points when it is available but difficult to find, confusing, incomplete, or cumbersome; and zero when it is absent or unusable. Do not treat an unverified feature as missing or award it full credit. Record what you could verify, identify the uncertainty, and explain how it affects the score. Base grades on the property-owner experience, not visual appearance alone.
Provide detailed improvement recommendations only for Gunnison County. Test Gunnison's full property-owner journey more thoroughly, including finding a record, understanding its value and property details, appealing, finding forms and exemptions, correcting information, and understanding property taxes. For each recommendation, identify the specific problem, the proposed change, why it would help an owner, and its estimated impact and implementation effort. Rank the recommendations by priority. Link to working examples from other Colorado counties when they offer a useful model.
Also provide dashboard-ready data as downloadable CSV files or a single JSON file. Include one record
per county with its name, assessor website URL, review date, seven category scores, total score, letter
grade, statewide rank, verification status, and brief summary. Provide separate records for Gunnison's
recommendations with the issue, proposed change, priority, expected owner impact, estimated effort, and
implementation status. Use consistent field names and values. Keep evidence links in the data, use
null for scores that cannot be verified, and do not combine multiple counties or
recommendations into a single record. Structure the files so they can be used to build a simple
interactive dashboard and updated after manual review.
If you cannot examine all 64 counties thoroughly, say so plainly. Do not present a partial sample as a definitive statewide ranking or percentile. Identify which counties were reviewed, label the comparison provisional, and leave unreviewed scores and ranks blank. Do not invent findings or links to fill gaps.