Domain-Citation Measurement

Shopigent AI GEO measures visibility through one honest path:

• Real-AI measurement (Growth/Pro): scheduled and manual, quota-limited runs that send buyer prompts through the selected, available engines: OpenAI web search and Perplexity Agent on Growth; Pro can also select Grok (xAI) web search; Claude (Anthropic) web search is available for manual runs only. A mention is a provider-attached citation of the store domain; retrieved search results and name matches are reported separately and never enter rates. Not a chat ranking.

The legacy web-search simulation is disabled.

Measurement does not guarantee future placement, recommendations, traffic, or sales.

What a Manual Run Reports

For each checked product, the app records:

• provider availability or failure
• whether the provider response attached a citation to your store domain
• the source/provider associated with the result
• a timestamped result that can be compared with later runs

A failed or unavailable provider is not saved as a successful measurement.

Monthly Limits

A competitive comparison consumes one manual quota unit, unchanged. Scheduled runs use a separate monthly counter and never consume manual quota.

Weekly Tracking

Weekly tracking defaults on for Growth and Pro. Turn it off in Settings to stop future scheduled questions; a job already queued still finishes and its reserved slot is not refunded. Free has no scheduled runs.

The same sample as manual runs builds up to 8 buyer questions from product titles. No question names your store domain. Weekly runs use the selected, available engines permitted by the plan; Pro can also select Grok (xAI). Claude (Anthropic) is manual-only and is excluded from weekly runs. API answers are not the signed-in chat products.

The scheduled window is Monday 03:00 UTC plus a stable shop stagger of up to 239 minutes. Crawl Monitor shows the next window, last scheduled run, and engines that completed. It does not promise that a cron is configured. Activation requires privacy deployment and owner approval.

The cap is 4 per UTC month, with no catch-up or fifth run in a five-Monday month. When used up, scheduling pauses until the next month; manual runs are unchanged.

Test Installs

Real provider measurement is disabled for test installs, including development/reviewer stores with test subscriptions, because provider calls incur real costs. Manual runs, scheduled runs, and competitive measurement do not call providers on these installs. A paid test subscription does not enable real measurement. No simulated success is substituted for the disabled provider call.

When to Run a Measurement

• after publishing a substantial product-description update
• after adding or changing important catalog products
• before and after a competitive comparison period
• when you want a new point-in-time provider response

Use the result as diagnostic evidence, not as a universal visibility score.