What is AI telling people about your organization?

Truthscope, by Firn, audits what ChatGPT, Claude, and Gemini say about your organization — and shows you what they get wrong and how to fix it.

Book an Intro Call See how it works ↓
by Firn
Dashboard
Overview
Accuracy
Coverage
Risk
Audit
Questions
Sources
Models
Settings
Settings
CT
City of Toronto
Active audit
City of Toronto Audit complete Overview By category By question Action items
Export Share report
AI accuracy report
How accurately do AI platforms answer questions about City of Toronto services?
111 questions · 2 models · March 2026
Average risk rate
27.9%
Wrong or partial answers
Critical errors
14.0%
Completely wrong answers
Model agreement
70%
78 of 111 same score
Highest risk
Community Crisis Response Fund
67% avg risk rate
Accuracy by model
ChatGPT37.8% risk
69 correct23 partial19 wrong
Gemini18.0% risk
91 correct8 partial12 wrong
Risk by category
CategoryChatGPTGeminiAvg
Child Care Subsidy75%25%50%
Noise60%40%50%
Snow Removal43%14%29%
On-Street Parking25%25%25%
Seniors Health20%0%10%
Evictions0%0%0%

Independent audit using publicly available data. Not affiliated with or endorsed by the City of Toronto.

The problem
AI gets 15–35% of institutional facts wrong. Your website visitors are already seeing it.

When someone asks AI about your fees, your deadlines, or your eligibility rules, it answers confidently, even when it's wrong. And you have no visibility into what it's telling them. Truthscope changes that.

What Truthscope does
Measure your exposure. See what to fix.
Accuracy
We score every AI response against your own published information — binary right or wrong. Every error comes with the source page it should match, so your team knows exactly what to correct.
Coverage
We audit across every topic your audience asks about — admissions, fees, policies, services, eligibility, deadlines, processes. Gaps tell you where to strengthen your content so AI has the right information to pull from.
Risk
Not all errors are equal. A wrong phone number is annoying. Wrong eligibility information locks people out of services they qualify for. We score severity so you know which errors to fix today and which can wait.
How it works
From ground truth to scored report in under a week
01
We map your ground truth
We identify the pages on your website that contain the specific facts your audience asks AI about — fees, deadlines, eligibility rules, processes. These become the source of truth every AI response is scored against.
02
We audit every major AI platform
We systematically query ChatGPT, Claude, and Gemini with the questions your audience actually asks — then score every answer against your ground truth. More models are being added.
03
You see exactly what's wrong
A clear report showing which topics AI gets wrong, how severe the errors are, and which areas carry the highest risk — with clear guidance on what to fix first.
Who it's for
Best for organizations with high-stakes, public-facing information
Universities
Admissions, aid, deadlines, program requirements, student services
Municipalities
Permits, benefits, recreation, housing, waste collection, deadlines
Service-heavy enterprises
Customer support, onboarding, eligibility, required steps, account access, and claims workflows

Every day, AI answers questions about your organization. You have no way to track what it's saying.

Get started
Find out where AI gets your most important questions wrong

We'll audit responses against your ground truth and show you which issues carry the highest risk — and what to fix first.

Book an Intro Call
FAQ
Common questions
How do you know an AI answer is wrong? +

We build a ground truth from your institution's own published information, including specific fees, deadlines, eligibility rules, and processes. Every AI response is scored against this ground truth. It's binary. The answer is either accurate or it isn't.

Which AI platforms do you test? +

ChatGPT, Claude, and Gemini. These are the three platforms that handle the majority of AI-generated search queries. More models are being added.

What do I actually receive? +

A comprehensive accuracy report scored by topic, by model, and by severity, with a risk heatmap and priority action items for the highest-consequence errors. You can drill down question by question to see exactly what each AI platform said and how it compares to your published information. With ongoing monthly monitoring, each audit tracks how AI accuracy shifts over time, detecting drift when models update and catching new errors when your policies change.

How is this different from checking AI myself? +

A single question tells you about a single answer. Truthscope tests systematically across dozens of questions, across every major model, scored against your published information, repeated monthly. You get a complete and ongoing picture of your institution's AI exposure, not a one-time spot-check.

Why can't the AI platforms just fix this? +

AI models are improving at general reasoning, but institutional accuracy is a different problem. Your information lives across dozens of web pages, dense PDFs, and policy documents that models struggle to parse correctly. Improvements in reasoning don't automatically translate to accuracy on your specific fees, deadlines, and eligibility rules. And as AI adoption grows, the impact of every error multiplies. More people are making real decisions based on what AI says about your institution, with no way for you to track it. For institutions where accuracy has real consequences, that distinction matters.