OhChat Review: Scored Against Five Criteria
OhChat scores 8.1 out of 10 against five criteria: persona control, memory span, content latitude, modality range and free-tier scope. It records the strongest persona and memory profile of the four apps assessed on this site, and the weakest modality range, since voice and images are limited and paid.
| Category | AI companion chat, 18+ |
|---|---|
| Criteria assessed | Five, unweighted |
| Composite | 8.1 out of 10 |
| Access | Web browser, desktop and mobile |
| Unpaid tier | Persona builder plus extended chat |
Bottom line
Two of the five criteria come out clearly ahead: persona control and memory span. Content latitude and free-tier scope are unremarkable, which in this category is a compliment. Modality range is the weak column. Whether 8.1 reads as high depends entirely on which of the five you were weighting.
What works
- A persona defined field by field rather than chosen from a catalogue
- Recall survives a closed session, not just the current thread
- Unpaid access covers ordinary use, not only setup
- No public profile or shared feed for a conversation to leak into
What to know
- Voice and image generation are limited and sit behind payment
- Browser only — nothing to install, and nothing available offline
- Long-lived memory carries forward early mistakes as readily as facts
The criteria, defined
Before anything is scored. Persona control is how much of a character you specify yourself. Memory span is how far back the system draws and whether recall survives a closed session. Content latitude is what the policy permits between adults and what it refuses. Modality range is text against text with voice and images. Free-tier scope is what unpaid access covers.
Each criterion is scored one to ten on its own terms, and the composite is the plain average with no weighting applied. The consequence is that the composite means little until you know which columns you would have weighted. A reader who cares mainly about generated media should expect to reach a different conclusion from the same five rows.
Criterion one: persona control
OhChat opens on an empty definition rather than a catalogue. Tone, backstory, interests and stated limits are separate entries, and the character that results is one nobody else is running. This front-loads effort: the first few minutes are configuration rather than conversation. Systems that hand over a ready-made persona reverse that trade, and for some readers the reversal is the point.
Across a long session the character holds its definition. Stated limits are not quietly dropped, and the tone set during setup does not drift toward a generic default after twenty exchanges. Editing a trait later takes effect in the next message rather than the next session, which is convenient and, when a character shifts mid-conversation, mildly disconcerting.
Criterion two: memory span
Recall here spans sessions rather than a single thread. Names, stated preferences and events described days earlier come back without being re-supplied, and they arrive inside ordinary sentences rather than as obvious callbacks. Set against thread-scoped systems, which answer well until the tab closes and then start over, this is the widest capability gap in the set.
The cost belongs on the record too. A long-span system carries forward whatever it retained early, including details stated loosely or in passing, and correcting an entrenched one takes more than a single message. A reader who prefers a clean slate each session will read this criterion as a drawback rather than a strength, using exactly the same evidence.
Criteria three and four: content latitude and modality range
Latitude is standard for the category. Adult themes between fictional characters are permitted, and the refusals are the ones that apply across every app covered here: no likeness of a real person, nothing involving minors, age confirmation before access. Nothing in the policy distinguishes OhChat from its alternatives in either direction, which is why the criterion scores mid-band rather than high.
Modality is where the score falls. The product is built around text, with voice on the paid tier and image generation thin next to apps designed around it. If a conversation without pictures is not the conversation you had in mind, this column outweighs the other four combined, and the composite score above becomes irrelevant to your decision.
Criterion five: free-tier scope
Unpaid access covers the persona builder in full and an extended run of ordinary text conversation, with no card requested at sign-up. That places it above sampler tiers that stop after a handful of exchanges, and below arrangements where text chat is effectively unmetered. Media features mark the paid boundary rather than basic conversation, which is the distinction worth noting.
This is the criterion to verify rather than read about, since it is also the one providers revise most often. Ten minutes inside the free tier will confirm or contradict this row directly, which is a better use of the time than weighing two comparison pages against one another.