ChatGPT
LiveAsked with web search on, so answers cite live URLs.
What we read · The URLs cited in the answer, in the order the engine lists them.
Default model · gpt-5.6-luna (8 credits) — a stronger model is selectable per call.
Coverage · engines
Five engines, reported one by one.
The same question returns different answers — and different citations — on every AI engine. Geondex asks each one the same question and reports each answer on its own, so a win on ChatGPT never hides a blind spot on Perplexity.
Claude
LiveAsked with web search on, so answers draw on live sources.
What we read · The sources its answer actually cites — a page it only searched and never cited is dropped.
Default model · claude-sonnet-5-5 (15 credits) — a stronger model is selectable per call.
Perplexity
LiveAn answer engine — every reply ships with a numbered source list.
What we read · That numbered source list, as returned by the engine.
Default model · sonar (2 credits)
Gemini
LiveAsked with Google Search grounding on.
What we read · The grounding sources behind the answer, resolved to the real page URLs.
Default model · gemini-3.5-flash (16 credits)
Grok
LiveAsked with web search and X search on, so it can cite live posts alongside pages.
What we read · The cited-span annotations in the answer — a source it only saw and never cited is dropped.
Default model · grok-4.3 (16 credits)
Copilot
Not yetMicrosoft's answer engine, built on the Bing index.
What we read · Not measured yet.
Why it matters
Same query. Different answers.
Each engine searches with its own provider and ranks sources its own way. None of them publishes how. A page that gets cited #1 on one engine can be invisible on another. The chart is an illustration of that, not a measurement.
That's why Geondex never averages engines into one number. You see each engine on its own, and what changed since your last check — so you can see which engines moved after a change you made.
Read the GEO field guide →Illustration · made-up engines, not measured
authorityfreshnessstructure
Engine A
Authority heaviest
Engine B
Structure heaviest
Engine C
Freshness counts more
All five engines are available on every account, including Free — each runs the model its own consumer app uses by default (or a fast model from the same provider where that one is too slow), at that model's credits, so all five defaults together cost 57 credits. Pick a stronger model where one is offered and it costs more. There is no schedule: a check runs when your agent asks and is stored so the next one shows the change.