Skip to main content
Human guide: work with your agent → Run one bounded DataForSEO search request and inspect its actual result before building a recurring research workflow.

​
Before you start

Use an authorized Deepline workspace. Inspect the action’s supported access mode and applicable price. A provider connection does not make every action free or establish that every access mode is supported.

​
1. Discover the actual operation

The catalog contains specific operations, not a universal dataforseo_serp_google or dataforseo_keywords_data tool. Read the returned schema and defaults.

​
2. Define one small query

This example requests one keyword with explicit location and language:
Inspect available location and language values before changing the scope. Keep the query, location, language, and collection time with the research result.

​
3. Execute only the reviewed request

This command starts provider-backed work and can incur usage:
Preserve its execution ID and returned recovery commands. Do not resubmit a new request merely because a caller stops waiting.

​
4. Check source status and result scope

Read the action’s declared output schema and returned status evidence. Separate a completed source task with no matches from an execution failure. Do not infer a competitor’s conversions or spend from search placement.
The result corresponds to the reviewed query and geographic/language scope. Missing or failed source evidence remains explicit.
Use the action reference for current fields. Read the exact billing command returned by execution. A missing charge amount is not a zero-charge guarantee.

​
5. Make the method repeatable

Author a checked Play when you need several queries, comparison rules, or a schedule. Retain scope and observation time across runs. See Write Plays and tool execution recovery. Ask your agent which source status explains an empty result before retrying.