Skip to content

Distinguish tool failure from a negative research result

Keep authentication, timeout and unavailable-analysis states separate from substantive evidence about the topic so operational gaps do not become false findings.

Guides

Did the review find nothing or fail to establish availability?

A research request about public battery recycling discussions may time out or return partial coverage. That is not the same as a completed review finding no qualifying material. Operational status is part of the evidence boundary, and the final summary should preserve the distinction rather than translating every empty response into no discussion exists about the topic.

Example research prompt
Classify supplied battery recycling research responses by completed-empty, partial and operational failure states. Preserve useful records and use only the documented recovery for each status.
01 / Guides

Read the returned status

Inspect the actual error or coverage information before interpreting content. Include partial, unavailable and completed-empty states separately. Exclude a substantive negative claim from authentication failure or timeout; those states describe the research operation, not the presence or absence of relevant public material in the wider world or within a source that was never successfully reviewed.

02 / Guides

Choose the documented recovery

Follow the returned guidance appropriate to the tool and error, such as honoring Retry-After for rate limiting. Do not blindly retry analysis_unavailable when the documentation says not to retry. Keep authentication recovery with the authorized user workflow rather than expanding credentials or permissions solely to make a research page’s example appear complete or successfully verified.

03 / Guides

Preserve useful partial evidence

Retain supported records with their coverage limits when a response is partial. Avoid claiming exact full-set counts from the visible subset. If an operational gap blocks a required comparison, mark that component unresolved instead of substituting unrelated material or assigning a zero value to the unavailable side of the table to maintain a clean-looking output.

04 / Guides

Deliver status-aware findings

State what was reviewed, what failed and which conclusion remains unsupported because of the failure. A methods example should not imply that recovery was actually attempted or successful. If the task cannot continue within authorized access and available tools, report the blocker directly rather than converting an infrastructure limitation into a public claim about a quiet topic or nonexistent audience.

Keep the boundaries clear.

  • Operational failure describes the research attempt, not the underlying topic’s absence.
  • Prompts are examples, not observed results; platform badges do not guarantee access or complete coverage.

Start with a read-only check: ask your agent to report actual source coverage before collecting records or exporting an audience.

Primary references

Documentation reviewed 2026-11-13. Product capabilities, client availability and plans can change. Third-party marks identify their products, not an affiliation.

Know what social media shows about you.
Start with a question.

Get started
montage