Build Your Own Report

• ChatGPT
  Model version

  • GPT-6 Astrameasured costNewest OpenAI flagship; two to seven searches an answer.
  • GPT-6.1 Solmeasured costThe ChatGPT the $180 three-engine report answers on; two to five searches an answer. Newer than the model the ChatGPT app serves, and the report says so.
  • GPT-6 Solmeasured costThe 6 family's everyday rung; half of 5.6 Sol's list, three to eight searches an answer.
  • GPT-6 Lunameasured costThe fast 6 rung; about a third of 6 Sol's bill.
  • GPT-5.6 Solmeasured costThe ChatGPT the consumer app serves by default; twice 6.1 Sol's bill an answer.
  • GPT-5.6 Terrameasured costMid rung of the 5.6 family; a little under half of Sol's bill.
  • GPT-5.6 Lunameasured costThe fast 5.6 rung; about a seventh of Sol's bill.
  • GPT-5.5measured costPrevious generation flagship; six to nine searches an answer.
  • GPT-5not yet availableTwo generations back. Not sold for now: it failed to return an answer on two of every three of our daily test calls.
  • GPT-4o minimeasured costWhat the free scorecard runs; one search an answer.
• Claude
  Model version

  • Claude Fable 5.1measured costThe frontier Claude: Anthropic's largest model, at the top of its price list.
  • Claude Opus 5.5measured costThe newest Opus, at two fifths of Fable's bill on the same search.
  • Claude Sonnet 5.5measured costThe Claude the $180 three-engine report answers on; a fifth of Fable's bill.
  • Claude Haiku 5.5measured costThe base Claude, about a fiftieth of Fable's bill; Anthropic lists it at a tenth of Haiku 4.5's price.
• Gemini
  Model version

  • Gemini 3.8 Flashmeasured costThe Gemini on every report, the free scorecard included; the same grounding fee as every 3.x Flash.
  • Gemini 3.7 Flashmeasured costOne back from the newest Flash; same fee, same bill.
  • Gemini 3.6 Flashmeasured costTwo back from the newest Flash; same fee, same bill.
  • Gemini 3.5 Flashmeasured costPrevious Flash; its tokens list above the rate 3.6 to 3.8 carry until the end of 2026.
  • Gemini 2.5 Flashmeasured costOlder family; Google bills its grounding at two and a half times the 3.x fee.
• Perplexity
  Model version

  • Sonarmeasured costThe vendor's fast preset: one search step, the shortest Perplexity answer.
  • Sonar Promeasured costThe Perplexity every all-engine tier answers on, served as the vendor's low preset: up to five search steps an answer.
  • Sonar Reasoning Promeasured costThinks before it answers, served as the vendor's medium preset; about half a minute and $0.02 a call, over twice Sonar's bill.
  • Sonar Deep Researchnot yet availableThe vendor's high preset. Forty searches and a minute or more an answer on the old API; one call on the new one answered in 17 seconds. Not sold until it answers whole every time.
• Grok
  Model version

  • Grok 4.7measured costNewest Grok, on 4.6's price list; it reads more search results an answer than any other Grok.
  • Grok 4.6measured costOne back from the newest Grok, on the same price list; reads fewer search results an answer than 4.7.
  • Grok 4.5measured costTwo back; nearly as heavy on search results as 4.6.
  • Grok 4.3measured costThe Grok every all-engine tier answers on.
  • Grok 4.20 Reasoningmeasured costReasons before it answers; reads about twice the search results 4.3 does.
  • Grok 4.20measured costSame model, no reasoning pass; reads about twice the search results 4.3 does.
• Google AI Overviews and AI Mode

  Both of Google's answer surfaces, on the same questions your engines answer. Every capture keeps whether an Overview showed, who it cited, and the sentence Google took from them.

  Captured on

  desktopmobile

  Asked from the United States, in English. The receipt prints the country, the language and the device that ran.

Every stored answer already names every brand the assistant mentioned, not only yours. A competitor report reads those same answers with the other name as the subject: their mention rate, their share of voice, where they were recommended and where they were passed over, on the same questions, engines and model versions. Nothing is asked twice and nobody is contacted. It says what the engines said, which is not the same as what is true, and it carries a line naming who it was measured for. Matching is by the name as typed, so spell it the way the assistants do.

Name a competitor in step 3 and it appears here, priced at what your report costs, with a maximum of $50.

One structured pass over your finished run: positioning, the 30-day quick wins, content briefs tied verbatim to the questions you lost, the third-party pages worth being on, the technical checklist, and how to measure whether it moved. It reads the same stored answers the report does and it never promises a ranking. The report's own prioritized fix list stays in the base price; this is the dated plan built on top of it, and there are three ways to have it.

Would you rather we made the changes? We implement the plan for you at our agency rates: the content, the placements, the schema, and the proof that it moved lands in your next report. Book a call and we'll quote it.