Method
How we test and report
Each review gives one task to several products a small team can buy, with ChatGPT alongside them, and reports what the recorded outputs did.
Run states
Every attempt is sorted into one state: tested with output, tested with no output, unreachable, or not tested. If a product was set up and given the task but produced no deliverable within the test window, that counts as a failure on that task and is scored 0, with its state shown (for example “output incomplete” or “still waiting”). Gaps on our side are never scored as product failures: a task we did not run, a product we could not reach, or records we do not have.
Strength of evidence
We use one of four labels: preliminary comparison, better in this sample, stable recommendation, or not enough evidence to separate. All current reviews are preliminary.
Scores, composites and rankings
Per-task results are always shown next to any total. A composite appears only with its weights stated (the AI-visibility review weights questions 30%, diagnosis 30% and actions 40%), and every place is numbered; neighbours that a small scoring change would reorder are marked “could swap”. The marketing-copy review reports four questions separately, with no total. The support-agent review shows evidence status only.
Who scored
Scoring in the current reviews was AI-assisted. Each article says which system judged which outputs, and when a rule was added after testing.
Updates and re-tests
We keep every review up to date. When a major new model is released, we re-run the tests and update the results; new evidence and corrections are added as they arrive. Each review lists every change with its date in its Updates section.
What we do not measure yet
Business outcomes (visibility, replies, conversions, sales), cost and staff time were not measured in these reviews. Each article says so next to its results.
Commercial relationships
analyst.ai has no commercial relationship with any product it reviews: no sponsorship, paid placement, affiliate links, or vendor-provided accounts or credits. Each review states this for its vendors.
Method and disclosures in each review
- AI visibility audit — Preliminary comparison, Review date 2026-10-11
- Marketing copy — Preliminary comparison, Review date 2026-10-11
- AI growth reports — Evidence status · no scores, Review date 2026-10-13
- AI ad creative — Evidence status · no scores, Review date 2026-10-13
- Cold sales emails — Preliminary comparison, Review date 2026-10-11
- Support handoff — Preliminary comparison, Review date 2026-10-11