Compare LLMs

The four LLM columns

Compare puts one prompt against the four engines that write prose — ChatGPT, Perplexity, Claude and Gemini — and shows the latest completed run for each. The engine names are model-specific, so you're reading Perplexity Sonar, Claude Sonnet, GPT and Gemini Pro rather than a brand badge.

Google and Reddit aren't here by design: Google is the search channel and its results sit on the prompt's own detail, and Reddit appears as thread citations.

What your plan fills

Which columns carry data depends on the engines your plan covers. On a three-engine plan the ChatGPT and Perplexity cards fill while Claude and Gemini stay empty until you move up.

Reading a card

Each engine gets a summary chip — mentioned (with a count when you're named more than once), not mentioned, or no data — plus the number of citations and competitors in that answer, the brand variants the model used for you, and a sentiment chip with its score. The answer itself highlights your brand in the accent tint and competitors in neutral grey. Sources are listed under each answer rather than as inline markers.

Cards can be dragged into the order you think in, and that order is remembered.

What to look for

• Coverage gaps — an engine that names you when the others don't, or the reverse.
• Framing — whether the description is current, and what it credits you for.
• Sources — which pages each engine leaned on, and whether any of them are yours.
• Naming — the variants models use for your brand, which is often where an alias is missing.

Comparing runs

Compare shows the latest run only. To see what changed between two runs, open the prompt's own page and use the Latest / Previous / Diff control there.