View-to-Favorite Ratios as a Quality Signal
Every character card on the trends page prints two counters, and the more useful number is the one you have to divide yourself. Eri shows 24.3M views against 7.1K favourites, a ratio of about 0.029 percent. Eva shows 19.1M against 2.8K, or about 0.015 percent. Reach says how many people arrived; the ratio hints at how many decided the character was worth keeping.
Key takeaways
- Ratios computed from the published counters: 7.1K against 24.3M is about 0.029 percent, 2.8K against 19.1M is about 0.015 percent, and 4.7K against 18.2M is about 0.026 percent.
- A view is an arrival, a favourite is a decision to save, so the ratio behaves more like a retention clue than a popularity score.
- The chart of 20 Best NSFW AI characters is refreshed every Thursday, which gives any ratio you calculate a shelf life of roughly one week.
- Illustrative habit arithmetic: four sessions a week across four weeks is 16 sessions, worth 4.0 hours at 15 minutes each and 16.0 hours at 60 minutes each.
What each counter actually records
A view is the cheapest action on the site. The card has to appear somewhere a person is already browsing and get one click. A favourite costs more: the reader has to decide that this character is worth returning to, which usually means they expect to open it again. That asymmetry is why the ratio matters more than the raw total.
Views scale with placement. A card near the top of a trending list collects them simply by being seen, and a card that appears in a search result or a ranking improves its count with every new reader. Favourites scale with intent. They accumulate slowly and they are harder to inflate with position alone, so a card with a modest view count and a strong save rate is telling you something a card with a huge view count cannot.
Neither number is a verdict. Both are card counters maintained by the site, read here on 22 Sept 2026, and both keep moving as the weekly chart is rewritten.
| Card | Views | Favourites | Favourites per view | Plain reading |
|---|---|---|---|---|
| Eri | 24.3M | 7.1K | About 0.029 percent | Widest reach on the board with an above-average save rate |
| Eva | 19.1M | 2.8K | About 0.015 percent | Broad curiosity, weaker signal that readers come back |
| Takara | 18.2M | 4.7K | About 0.026 percent | Roughly three quarters of Eri's reach with a comparable save rate |
| Board median | 18.2M to 24.3M | 2.8K to 7.1K | About 0.015 to 0.029 percent | A narrow band, which makes outliers easier to spot |
Worked example: turning two counters into one decision
Start with the division. Eri: 7,100 saved against 24,300,000 opened is 0.000292, which is 0.029 percent. Eva: 2,800 against 19,100,000 gives 0.000147, or 0.015 percent. Takara: 4,700 against 18,200,000 gives 0.000258, or 0.026 percent.
Then invert it, because one save per how many views is easier to picture than a decimal. At Eri's rate that is one favourite per 3,423 views. At Takara's it is one per 3,873. At Eva's it is one per 6,821, which is the number that separates the table's top row from its bottom row in practical terms.
Now attach the ratio to a habit. Take an illustrative reader who opens the site four times a week and reads for 15 minutes each time. That is 4 sessions a week, so 16 sessions over four weeks, at 60 minutes per hour: 15 minutes times 16 is 240 minutes, which is 4.0 hours a month. At 30 minutes per session the same 16 sessions total 480 minutes, or 8.0 hours. At 60 minutes per session they total 960 minutes, or 16.0 hours. Those session figures are an illustrated assumption about one person's habit, not platform data; the counters above are the real thing.
What the ratio does not replace
A ratio is a summary, and summaries hide the shape of the data underneath. Two cards can both land on 0.026 percent with completely different histories: one collected saves steadily over months, the other collected almost all of them in a single week after appearing at the top of the board. The figure is identical, the trajectory is not, and only the second card asks you to wait a week before deciding anything.
The ratio also does not measure what a reader means by quality. A favourite records that somebody chose to save a character, not that the conversation held up afterwards. A saved character that is never opened again still sits in the count. Treat the ratio as a shortlist of candidates rather than a certificate.
Where it works well is comparison inside one page. On Crushon, the trends board places a fixed set of entries side by side with the same two counters attached to each, so the division is fair across the row. That fairness disappears the moment you compare a card on the board with one you stumbled on three pages deep in a filtered search, because the second card has never had the same exposure.
One more limitation is worth stating plainly. Because the counters are read at a moment in time, every ratio in this article describes 22 Sept 2026. A character's standing can change within a week, and the honest response is to recalculate rather than to quote the old figure as though it were permanent.
Three ways the ratio misleads
The first trap is age. A card published last week has had far fewer chances to collect favourites than one that has been in rotation for months, so a low ratio on a new entry says more about its date than its quality. Check the weekly chart at https://crushon.ai/trends/nsfw_ai before treating any ratio as a judgement, because the list is rewritten every Thursday and new entries arrive mid-cycle.
The second trap is exposure. Views arrive from the homepage, from character search and from the weekly rankings, and each route brings a different kind of reader. A card surfaced to casual browsers will dilute its own ratio, which is not a failure of the character.
The third trap is treating favourites as steady. The counter only moves forward while readers keep saving. If a card leaves the front page, the ratio freezes while the view count keeps drifting, and the pair stops describing the same moment.
Frequently asked questions
Is a higher ratio always better?
No. A niche card can post a high ratio on a small view count, which is a strong fit for a narrow audience rather than a broad success. Compare ratios only between cards with similar reach.
How often should the counters be re-read?
Weekly matches the refresh rhythm of the chart. Reading them daily adds noise from the placement changes that happen inside a week, not signal.
Do favourites change anything visible on the card?
They are displayed alongside views as a public counter. What you do with the number is your own analysis, not a ranking the site promises to honour.
A five-minute weekly routine
Open the trends board, note the three cards with the highest view counts, and divide each favourite count by its view count. Write the three ratios down. Next week, repeat, and look only at which ratios moved. Cards whose ratios rise while their views hold are the ones worth opening yourself, and cards whose views spike while the ratio drops are being shown widely rather than being saved.
The routine takes about five minutes and replaces a vague impression of what is popular with three numbers you can defend. Keep the same three ratios in a note, and after a month you will have a small personal record of how the board behaves between Thursday refreshes.
Data and sources: card counters and the 20-character weekly chart on the trends page[1], with the weekly ranking schedule and saved-character search described on the search page[2]. Counters were read on 22 Sept 2026; session-hour figures are worked examples, not platform measurements. Written for adults 18+; read 22 Sept 2026.