Skip to main content

​
Direct answer

Run the same configured prompts across the providers and models you want to compare, then review the stored answers side by side. Classify recurring claims, sentiment, and omissions. Consistency means similarity within that selected dataset, not agreement across every possible response.

​
Data required

Use identical configured prompts, aligned run periods, provider and model identifiers, stored answers, topics or tags, sentiment or perception labels, and citation URLs. Keep the provider family distinct from the selected model.

​
Workflow

1

Create the comparison cohort

Choose prompts that can be run meaningfully across the selected providers. Use the same brand, topic, geography, and period filters.
2

Read answers side by side

Compare factual claims, positioning themes, sentiment, SELF and DIRECT mentions, and recorded citations.
3

Separate stable and isolated patterns

Mark themes repeated across several answers separately from one-off wording or provider-specific omissions.
4

Recheck material differences

Repeat important comparisons on later runs and verify factual claims against approved sources before escalating them.

​
Interpretation

Differences are observations, not proof that one provider is correct or biased. A provider is a service family; a model is a selectable model within that family. Changes in model, web access, prompt execution, or source availability can affect the comparison.

​
Possible next actions

  • If a factual discrepancy repeats, test a correction through accurate public source material.
  • If one model omits a positioning attribute, investigate the answers and citations before assuming a distribution problem.
  • If narratives are broadly consistent but outdated, prioritize updating the authoritative source of the outdated claim.

​
Limitations

Stored answers are samples from configured runs. They do not represent every provider output, user context, locale, or future model version. The REST answers endpoint is a separate data contract; confirm its fields and pagination before comparing exports.

Exports

Export the selected comparison cohort.

Answer Details

Build provider and model comparisons in Looker Studio.

Answers API

Read the separate REST answers contract.

Prioritize providers

Choose providers relevant to your audience.

What AI says

Review answer narratives directly.

Why a competitor is recommended

Investigate a provider-specific comparison.

Perception versus positioning

Compare observed themes with approved positioning.