← All toolsAnother direction / 01 Another direction / 02 Another direction / 03
The selection guideSource-backed research ↗
Analytics / Open-source causal GEO benchmark and AI citation experiment runner
OpenGEO
Open, reproducible experiments for measuring what makes AI answer engines cite a page
Causal GEO experimentationAI citation measurementOpen-source research reproducibility
Why it stands out
Paired design is narrower and more interpretable than a blended visibility score.
Source review: · Public-source research, not a hands-on test
↗ Best for
GEO researchers designing reproducible experiments.
! Know before you buy
Measures synthesis and citation after documents are handed to the engine; does not measure crawling, ranking, retrieval eligibility, or unaided discovery
01 / Editorial assessment
Is OpenGEO for you?
The AICiteKit take
OpenGEO is a strong methodological fit when a team needs to test one concrete page edit against a defined question and competitor set. The project publishes its paired design, corpus, harness, raw responses, analysis ledger, and uncertainty intervals, and it explicitly separates citation after document access from discovery and retrieval. Its published rounds are useful evidence for hypotheses, not universal ranking rules: effects depend on the page, prompt, competitors, models, and observation date. No independent customer-rating sample or external replication was found in this evidence pass. Choose OpenGEO for controlled measurement; choose a monitoring platform for continuous brand, prompt, competitor, and citation coverage.
A good fit if you are…
- GEO researchers designing reproducible experiments.
- Technical content teams testing a defined page revision.
- Agencies separating causal evidence from observational visibility scores.
- Developers wanting raw responses and a self-hosted workflow.
Consider another option if…
- Marketing teams needing an always-on visibility dashboard.
- Buyers needing crawlability, ranking, or unaided discovery measurement.
- Teams unable to control model/API budgets and experiment metadata.
- Anyone seeking guaranteed traffic, ranking, conversion, or revenue effects.
02 / Scope, not promises
Features & coverage
- Experiment design
Paired controlled comparison of an original page and an edited page against the same question and competitors, with randomized document order and repeated runs
Archived research · Official README and methodologyResearch excerpt · Paired citation experiments- Experiment outputs
Citation-rate change with 95% interval, per-engine results, charts, raw model responses, and a machine-readable findings ledger
Archived research · Official overview and repositoryResearch excerpt · Published findings and reproducibility- Measurement boundary
Measures synthesis and citation after documents are handed to the engine; does not measure crawling, ranking, retrieval eligibility, or unaided discovery
Archived research · Official READMEResearch excerpt · Retrieval is intentionally held constant- Independent evidence
No independent product review, representative customer-rating sample, or external replication found; GitHub activity is bounded adoption context only
Archived research · Evidence review checked 2026-11-01Research excerpt · What feedback does and does not establish
Paired citation experiments
Run the original and edited page through the same question, competitors, repeated trials, and analysis plan to estimate a conditional citation-rate change.
Published, inspectable findings
Review the public overview, findings page, methodology, raw-result structure, and machine-readable ledger rather than relying on an opaque composite score.
Causal boundary made explicit
Use the measurement to test a page edit after document access, then pair it with a crawler or monitoring tool if discovery, retrieval, or ongoing visibility is the real requirement.
A focused evaluation workflow
- Define the question, original page, edit, and real competitors.
- Run the documented dry run and record the project revision and estimated budget.
- Execute paired repeated trials with fixed metadata and a pre-specified analysis plan.
- Inspect per-engine citation rates, intervals, raw responses, and null results.
- Repeat on another page or unpublished question before generalizing.
03 / Commercial boundaries
Pricing & limits
Terms below reflect the 2026-11-01 source review, not a new live quote. Confirm currency, billing term, tax and required scope with the provider.
- Public cost signal
Free MIT software; README estimates about $0.30 per question at current prices, while model/API and hosting costs vary
Archived research · Official README checked 2026-11-01Research excerpt · Pricing and limits- License and data
MIT code; README identifies published data and results as CC BY 4.0
Archived research · Repository README and licenseResearch excerpt · Pricing and limits
Additional costs & usage boundaries
Include the usage components listed above in the total cost; they are not separate flat software subscriptions.
Your buying checklist
Get these answers first.
- Which model versions and answer surfaces are included in the round I am relying on?
- Does my question have enough headroom, and what run count and budget are justified?
- How will I record retrieval, region, prompt, page, corpus, and date metadata?
- How should I combine the experiment with crawlability and ongoing visibility monitoring?
- Can a second team reproduce the result from the published ledger and raw responses?
Pricing context & qualifications
Checked November 1, 2026: MIT software is free to run, but the documented API workflow incurs model/API costs. The README gives an approximate $0.30-per-question planning signal and provides a dry-run estimate; actual prices and budgets vary.
04 / The tradeoffs
Strengths & weaknesses
Reasons to shortlist
- Paired design is narrower and more interpretable than a blended visibility score.
- Public methodology, corpus, raw responses, and findings ledger improve inspectability.
- CLI exposes dry-run cost estimation and per-engine uncertainty intervals.
- Publishes nulls and caveats instead of only positive findings.
- MIT code can be inspected, self-hosted, and adapted.
Reasons to pause
- No independent customer-rating sample, third-party review, or external replication found.
- Document handoff holds retrieval constant and excludes discovery and ranking.
- Effect sizes depend on corpus, prompts, competitors, model versions, and date.
- Model/API and hosting costs are separate from the free software.
- Citation-rate change does not establish traffic, conversion, or revenue.
Measures synthesis and citation after documents are handed to the engine; does not measure crawling, ranking, retrieval eligibility, or unaided discovery
05 / Choose by the job
Alternatives to consider
Editorial fit comparisons—not performance rankings or claims that one option always wins.
06 / What we can substantiate
Reviews & evidence
No independent product-specific review sample or external replication was found; evidence is medium for the published implementation and experiment design, and low for generalizability or business outcomes.
AICiteKit assessment: public-source research, not a hands-on product test. Official materials establish stated scope; reviews are dated and may concern the broader platform. Neither establishes guaranteed citations, traffic or revenue.
Sources & original verification dates
| Source | Evidence type | Checked |
|---|---|---|
| OpenGEO official overview ↗ | official | |
| OpenGEO GitHub repository ↗ | official | |
| OpenGEO published findings ↗ | official | |
| OpenGEO Zenodo archive DOI ↗ | archive | |
| OpenGEO public GitHub activity ↗ | community |
Original review evidence & date context
Preserved from the published research. Dates, review counts and product-scope qualifications below have not been refreshed by this layout migration.
The official overview, GitHub repository, findings page, license, and Zenodo archive support identity, paired experiment design, published findings, CLI workflow, and licensing. GitHub activity is bounded community context only. No independent customer review, representative rating sample, or external replication was found, so generalization and business impact remain unproven.
Research notes & unresolved discrepancies
- GitHub activity and Zenodo support existence, preservation, and bounded community context; they are not independent efficacy reviews.
- The cited +48-point result is a project-published, setup-specific finding, not a promise for another site.
- No citation accuracy, traffic, conversion, ranking, or revenue claim is supported beyond the project's defined experiments.
Layout migration: October 4, 2026. The original MDX research is retained in the repository; this reorganization does not refresh its factual verification dates.