Open-source GEO detection

Measure product visibility in official AI Web answers.

Aloom collects answers and visible sources from the official Web interfaces of Doubao, DeepSeek, Yuanbao, and Qwen. Versioned prompt suites produce comparable reports without substituting model API output.

Provider sessions and browser profiles stay on the collector machine. Browser profile files are not uploaded.

Example product · Quick Scan Official-Web sample
Aloom GEO Score
78.5
illustrative
Mention rate
76%
analyzed samples
Completion
100%
72/72
ProviderMentionComplete
Doubao 78%18/18
DeepSeek 72%18/18
Yuanbao 83%18/18
Qwen 69%18/18
Measured on official Web Doubao DeepSeek Yuanbao Qwen Provider names and logos are trademarks of their respective owners. Aloom is not affiliated with or endorsed by these providers.
Illustrative sample · not client data

What a short Aloom report looks like.

This deliberately simple example shows the shape of an output, not a claim about any real brand.

Subject: example analytics product Scope: 4 official Web providers Series: baseline

Results reflect observed AI visibility for the selected time, providers, prompts, and modes. They do not represent every user scenario or guarantee future visibility.

VisibilitySeen on 3 of 4

Often mentioned, but not consistently cited.

The product appears in several broad recommendations, with uneven source visibility across providers.

Common contextGeneral category

Recognized mainly as a general analytics tool.

More specific positioning and comparison language appears less consistently in the sampled answers.

Source signalOfficial pages

Product and documentation pages are the clearest sources.

Public product facts are easier to verify when positioning, capabilities, and comparisons are explicit.

Possible next step Clarify the positioning and comparison pages, then rerun the same prompt pack after 30 days.

Repeatable detection, not a one-off screenshot.

Every formal series freezes its prompts, provider modes, sample depth, and checkpoints so results stay comparable as AI answers change.

01

Confirm the product facts

Scan the official product website, then confirm the brand, category, products, audiences, regions, and competitors.

02

Run a fixed detection pack

Use versioned prompts across selected providers and modes. Each baseline prompt starts in a fresh official-Web conversation.

03

Inspect every denominator

See completion, mention, recommendation, rank, sentiment, sources, competition, factuality, and stability by provider.

The product surface matters

APIs and official Web are different observation surfaces.

Wording, ranking, modes, and visible citations can differ. Aloom observes the official interfaces instead of substituting model API output for monitoring samples.

Observed, not simulated
Measurement detailOfficial WebAPI only
User-visible answerObservedDifferent surface
Visible sourcesCapturedNot equivalent
Provider modeVerifiedAPI-specific
Login sessionLocal onlyNot measured
Repeatable seriesVersionedPossible

Open-source core. Implementation help when needed.

Aloom Core is open source under Apache-2.0. Setup, private deployment, provider adaptation, and recurring report interpretation are available as project-based services.

Open source

Aloom Core

Free

Run it on your own macOS or Windows machine.

  • Official-Web collection
  • Versioned prompt suites
  • Provider and multi-cycle reports
  • Your provider accounts and API key
View self-hosting guide
Ongoing

Monitoring & reports

Quoted by cycle

Repeatable runs with a fixed configuration and human interpretation.

  • Comparable multi-cycle runs
  • Change and competitor review
  • Provider-adapter maintenance
  • Human report summaries
Discuss a report

What the report answers

See where your product appears and how that visibility changes.

Where is my product mentioned?

Review mentions, recommendations, rank, sentiment, and visible sources by provider, prompt, and mode, then inspect the original answer for context.

Are detection cycles directly comparable?

Each series fixes the prompt version, providers, modes, languages, and sampling definition. Multi-cycle reports show a Comparable trend and retain detailed metrics for every cycle.

What happens to failed samples?

Failed, blocked, and unattempted checkpoints remain in the denominator. Completion and visibility are reported separately so missing samples do not make the result look stronger.

What does the Aloom GEO Score mean?

It combines six measurement layers: visibility, evidence, factuality, competition, stability, and governance. Use it for comparisons under the same methodology, not as a guarantee of visibility or causality.

Establish your first comparable AI visibility baseline.

Follow the GitHub self-hosting guide, or ask about first-detection and deployment support. Use the first report as a baseline for later cycles.

Contact: z1250835369@gmail.com

★ Aloom is developed in the open. Star the repository to follow releases and provider updates. Star Aloom and follow updates ↗