Reading View and Favorite Counts on a Character Page
A view count measures how many people looked; a favourite count measures how many decided to keep. On the shelf read for this piece on 22 Sept 2026 the widest pair was 24.3M views against 7.1K favourites, a save rate of 0.029%. The second number is the informative one, and it is the one most readers scroll past.
Key takeaways
- Rate beats traffic: 7.1K favourites on 24.3M views is 0.029%, roughly one favourite for every 3,423 viewers of that card.
- The three cards read here range from 0.015% to 0.029%, so a single shelf can differ twofold in how well it turns attention into intent.
- A view counter grows with placement and age; a favourite counter grows only when a reader deliberately keeps the card.
- Read a card in three passes - description, tags, then ratio - and let the third pass break a tie, never lead.
What the two counters are actually measuring
A view counter is a distribution number. It rises when a card is placed high on a shelf, appears under a tag somebody searched for, or simply sits there long enough to collect traffic. A favourite counter is a decision number. Nobody favourites a card by scrolling past it; the action is deliberate, which is why favourites are counted in thousands while views are counted in millions.
That asymmetry is the whole point. Comparing 7.1K with 24.3M puts a small intentional subset beside a very large mostly accidental one. The ratio between them is steadier than either figure on its own: it does not care how old the card is, how prominent it was last Thursday, or how often the same reader was served the same entry twice.
There is one reading trap. Both counters are cumulative. An older card that has been on a shelf for months will always out-total one that appeared this week, even when the newer entry is holding attention better. A ratio dampens that effect without erasing it, because the audience of a card also changes as it gets older.
Three readings to place side by side
Three readings, then the shape of the shelf they came from. The list is the Crushon weekly chart of 20 NSFW AI characters, and its rankings refresh every Thursday, so all three counters were read on the same day from the same page. Eri showed 24.3M views and 7.1K favourites, which divides to 0.029%. Takara showed 18.2M and 4.7K, or 0.026%. Eva showed 19.1M and 2.8K, or about 0.015%. Three cards, one week, a twofold spread in rate.
Notice what the view totals do by comparison. They sit within 1.34 times of one another, from 18.2M to 24.3M, so raw traffic would suggest the three cards are near enough identical. Only the rate separates them. That is the argument for doing the division before you decide anything, and for treating a prominently placed card with a thin rate as a card that was seen more than it was wanted.
| Signal on the card | Best used for | Where it misleads |
|---|---|---|
| View count | Judging reach and recalling where the card sat on a shelf | Grows with age and placement as much as with interest |
| Favourite count | Spotting entries that other readers chose to keep | Rewards early exposure rather than present-day fit |
| Favourite-to-view ratio | Comparing cards of different ages on one scale | Hides session length, repeat visits and how long anyone stayed |
| Tag row such as Unfiltered, Submissive, Fictional, OC | Checking whether a card matches the mood you brought with you | Describes the card's framing, not the quality of its writing |
| Creator credit, shown as "by ..." | Following a creator whose style already works for you | Says nothing about one individual card in isolation |
Worked example: from a shelf rate to your own six weeks
Run the division in order. 7.1K divided by 24.3M is 0.000292, which is 0.029%, or one favourite per 3,423 views. 4.7K divided by 18.2M is 0.000258, or 0.026%, one per 3,872. 2.8K divided by 19.1M is 0.000147, or about 0.015%, one per 6,821. Same shelf, same reading date, a twofold difference in how much of the audience kept the card.
Now hold that against the part of the ledger the counters cannot see. Take the ramp plotted below, which is illustrated arithmetic rather than platform data: 3 plus 4 plus 5 plus 6 plus 7 plus 9 sessions over six weeks, a cumulative 34, an average of 5.7 a week. At twelve minutes a session that is 408 minutes, or 6.8 hours. You are one reader inside a 24.3M-view audience, and your handful of favourites is part of the 7.1K.
Both halves matter, and they are of different kinds. The platform side is measured and published on the cards. The reader side is an assumption built only to show the arithmetic, not a claim about how anyone actually behaves inside the product.
How to read a card in three passes
Pass one is the description. Read the first two sentences and ask whether you can predict the tone of the opening exchange; if you cannot, the description is doing decoration rather than work. Pass two is the tag row and the creator credit, which tell you what the card is framed as and who wrote it. Pass three is the arithmetic, and it is the shortest pass of the three.
Then read the card where you found it, because placement drives views. Open the nsfw ai chat shelf and note where the entry sits, what sits above it and whether the neighbours carry the same tags. A card converting at 0.029% from the middle of that list is doing something a top-placed card reading 0.015% is not, and that difference is worth more than the traffic gap.
Keep the arithmetic small enough to finish. Three or four ratios, then one decision. If two cards sit within about 0.005% of each other, the counters have stopped being decisive and the description should carry the choice instead. When you want a wider sample than one ranked shelf, the character search page pulls cards by tag rather than by position, which is a better base for ratio work because placement no longer flatters the same few entries.
Frequently asked questions
Which counter should a new reader check first?
The ratio, because it normalises two cumulative numbers that grow at different speeds. On the cards read for this piece the rate ran from 0.015% to 0.029%.
Do favourites measure quality?
Not directly. They count how many readers decided to keep a card, which mixes writing quality with placement, tags and how long the card has been visible.
Why not simply pick the card with the most views?
Because traffic mostly reports shelf placement. The three cards read here had view totals within 1.34 times of each other while their rates differed by a factor of two.
Do the division once, then stop
Pick three cards from a shelf, divide favourites by views, and write the three rates down. If the descriptions are of similar quality, take the higher rate; if they are not, take the better description and ignore the counters entirely. Then put the numbers away for a month, because a rate read weekly starts to feel like a scoreboard rather than a filter. Published counters on the platform are one input into a decision; the other input is a month of your own sessions, which no card can show you.
Figures come from the view and favourite counters printed on trending character cards at crushon.ai and were read 22 Sept 2026[1]. The six-week session ramp is own arithmetic, labelled as an illustrative usage profile rather than platform data. Per the platform's own statement its content is AI-generated text and depicts no real person, so no counter here describes images or video. Written for an adult readership; the platform is intended for people aged 18 and over.