The AI BlackboxEnds Here
[ 01 ]Independent GEO oversight

Your GEO provider should not grade its own work.

Subscribe to independent observations of AI visibility and recommendation rank, then give your GEO provider objective TDR metrics to improve.

[ 02 ]The conflict in GEO measurement

The provider chooses the queries. Runs the test. Reports the win. Who verifies the result?

OptimiserData ownerScorekeeper= one conflicted system
[ 03 ]Research archive

Claims tested in public.

Investigations, public methods and worked audits designed for brands, investors and journalists who need evidence they can trust.

3 documents
INV-001August 2026
Investigation

Pxxxx Case Study: Could These Results Occur by Chance?

A worked statistical examination of a GEO provider's published case-study claim, using demonstration data and a pre-specified null model.

Claim testp < .001
PDF / 6 pp.
MTH-001Version 1.0
Methodology

A Reproducible Standard for GEO Visibility

Our public framework for measuring AI appearance rate and recommendation rank before and after a GEO engagement.

ProtocolPre / post
PDF / 6 pp.
AUD-001Sample report
Audit sample

Independent GEO Audit: A Worked Example

A redacted example showing the evidence, uncertainty and conclusions a brand receives from an independent effectiveness audit.

VerdictNot proven
PDF / 6 pp.
Notice

Initial documents serve as methodological proofs-of-concept. Production measurement programmes require verified client data.

[ 04 ]How we produce independent GEO data

Defined for your audience. Observed continuously. Verified statistically.

For each subscribed brand, we define the target audience, markets, AI models and core metrics. We monitor continuously and version the prompt test set as our research identifies changes in audience behaviour, preserving comparability over time.

01

Define the Measurement

Specify the brand, target audience, markets, AI models, observation window and metrics before monitoring begins.

02

Observe Continuously

Track AI appearance rate and recommendation rank over time, versioning test prompts as audience behaviour evolves.

03

Validate Statistically

Check anomalies, uncertainty and trend reliability across repeated observations before reporting a change.

Always public

Investigation reports and provider research

Public methodology and measurement standards

Long-term GEO industry observations

Longitudinal subscription data

Recurring AI appearance-rate observations

Recommendation-rank tracking across agreed models

Before/after trends with statistical uncertainty

Data subscriptions

Give your GEO provider an independent target.

Subscribe to recurring TDR observations of AI appearance rate and recommendation rank. Share these objective metrics with your GEO provider, so optimisation is accountable to data the provider does not control.

Explore Data Subscriptions
Subscription dataTDR / GEO
01Baseline

Current AI appearance and recommendation rank

02Monitoring

Recurring observations across agreed models and markets

03Trends

Changes over time with statistical uncertainty

04Reporting

Independent data for brand and provider teams

Independence pledgeWe will never sell GEO optimisation services.
[ 05 ]Contact The Dragonfly Research

Engage with us.

GEO measurement subscriptions

Independent GEO data, continuously.

Subscribe to ongoing measurement of your brand across ChatGPT, Gemini, Claude, DeepSeek and more. Track TDR metrics over time and use them as the objective performance target for your GEO programme. Your provider can optimise against the indicators, but cannot grade or rewrite them. Independent audits remain available when needed.

Recurring AI appearance-rate trackingRecommendation-rank monitoringBefore/after trend analysisRegular data delivery for brand and provider teams
Contact usEnquiry
Service required

This opens a prepared email to our sales team.

The Dragonfly Research

The GEO Watch.

Independent research for brands, investors and journalists who need GEO evidence without a commercial interest in the outcome.

Enter the research archive