It tells you how many people took the text, and almost nothing else. In a 20-prompt snapshot of the Faymas marketplace feed taken on 30 September 2026, the most-copied prompt on the page (76 copies) rendered at position 19 of 20, and feed position correlated with copies at ρ = 0.04 — statistically indistinguishable from no relationship at all. Read the copy count as a receipt of demand that already happened, not as a lever you can pull to get seen.
This guide reports measurements from a single direct fetch of the Faymas marketplace feed on 30 September 2026. No generation tests were run, and no platform-wide claims are made.
The copy count is the only counter on the card that moves
A Faymas prompt card carries four numbers in its data: copies, views, likes and saves. Only one of them is worth reading. In our snapshot, views read 0 on all twenty cards. There is no view counter rendered on the card front at all — the field exists in the page payload and stays empty.
Likes barely register either. Across the same twenty prompts, likes totalled 11. Twelve of the twenty cards had zero likes. The single most-copied prompt on the page had zero likes and ten saves.
Saves sit in the middle: 36 across the sample, with seven cards at zero. That is three times the like count on the same cards. If you want an early read on a prompt you just published, saves respond before likes do.
Copies totalled 423, and only one card sat at zero. So the hierarchy is blunt: copies is the signal, saves is the early warning, likes is noise, views is not wired up.
The distribution is steep — but the cliff is shorter than it looks
Sorted high to low, the twenty copy counts were: 76, 64, 50, 38, 32, 27, 24, 18, 17, 17, 11, 11, 10, 9, 6, 6, 3, 3, 1, 0.
| Band | Copies per prompt | Prompts | Share of all copies |
|---|---|---|---|
| Top (rank 1–2) | 76, 64 | 2 | 33.1% |
| Upper (rank 3–4) | 50, 38 | 2 | 20.8% |
| Middle (rank 5–10) | 32, 27, 24, 18, 17, 17 | 6 | 31.9% |
| Lower (rank 11–20) | 11, 11, 10, 9, 6, 6, 3, 3, 1, 0 | 10 | 14.2% |
Two prompts out of twenty took a third of all copies. The bottom half of the sample took 14.2%. That is a real skew.
But the number that matters for a new creator is not the top. It is the median: 14. The mean is 21.1, and the mean lies — it is dragged upward by two cards. Quote 21 as "typical" and you overstate the middle of the market by roughly half.
The top card is 5.4× the median. Not 76×. The gap between a forgettable prompt and the best prompt on the page is five-fold, which is a distance you can close. The genuinely bad outcome is the zero, and exactly one card in twenty hit it.
Feed position does not follow copies
This is the part that changes how you should behave. We compared where each prompt actually rendered on the page against each of its numbers, using Spearman rank correlation across the twenty positions.
| Measured against on-screen position | Correlation | Reading |
|---|---|---|
| Copies | 0.036 | None |
| Saves | 0.017 | None |
| Age (hours since posting) | −0.132 | Weak |
| Payload score | 0.203 | Weak |
Nothing predicts placement. The concrete cases are more legible than the coefficients:
- The 76-copy prompt — also the most-saved, with 10 — rendered 19th of 20.
- The 64-copy prompt rendered 16th.
- A 0-copy prompt, posted 48 minutes before we fetched the page, rendered 10th.
- A 38-copy prompt rendered 2nd.
The page payload arrives with its sort parameter empty and a seed value of 0 alongside the prompt list. Our inference — and we flag it as inference, because we found no published description of the ordering rule — is that the default feed is shuffled rather than ranked. Either way, the measured outcome is the same: being good does not put you higher.
There is a score field, and it is not a popularity score
Each prompt in the payload carries a numeric score, ranging from 1.26 to 11.53 in our sample. It is tempting to read that as a ranking. It is not one.
Score against copies: ρ = −0.05. No relationship. Score against age: ρ = −0.71. Strong, and negative — newer prompts score higher.
The examples make it plain:
- A prompt with 0 copies, posted 0.8 hours before the fetch, scored 7.83.
- A prompt with 76 copies, posted 22.6 hours earlier, scored 5.66.
- A prompt with 38 copies, posted 24.3 hours earlier, scored 2.73 — near the bottom of the sample.
Whatever that score drives, it is a recency-weighted value, not a quality rank. We could not confirm what consumes it; it does not order the feed we measured, and the league leaderboard page returned no standings when we fetched it. Treat the explanation as unconfirmed.
Two more things the snapshot shows
Category concentration. Seventeen of the twenty prompts were filed under photography. Two were social media, one was wallpapers. The marketplace is not yet broad; a prompt in a thin category faces less competition for attention and a smaller audience for it.
Model split. Fifteen cards carried the chatgpt tag, five carried gemini. The higher-copy cards were disproportionately on the minority tag — the gemini-tagged group had a median of 27 copies against 11 for chatgpt-tagged — but five data points cannot carry that claim, so we report it and do not build on it.
What to do with this if you just published
Five short rules follow from the measurements above.
- Target the median, not the top. Fourteen copies is the midpoint of a live Faymas feed. Twenty is already above the middle.
- Do not read a low position as failure. Position carried no information about copies in our sample. A prompt at slot 19 was the most-copied one on the page.
- Judge a prompt at 24 hours, not at 1 hour. Every counter except the recency-weighted score needs time to accumulate.
- Watch saves, not likes. Saves moved on thirteen of twenty cards; likes moved on eight. Likes are too sparse to steer by.
- Post when your audience is awake. If the score is recency-weighted — inferred, not confirmed — then publish timing is one of the few inputs you control that plausibly affects visibility.
Where these numbers come from, and what they cannot tell you
Method: a direct fetch of the Faymas marketplace home page on 30 September 2026 at approximately 09:30 UTC+8. Twenty prompt cards were on the page. Copies, views, likes, saves, model tag, category, payload score and creation timestamp were read from the page's embedded data; on-screen order was read from the card markup. Correlations are Spearman rank coefficients computed over those twenty observations.
What this snapshot cannot support:
- Platform-wide claims. Twenty cards is one screen of one feed, not the catalogue.
- Earnings. We found no published rate converting copies into money, so none is asserted here.
- Ordering rules under other conditions. We measured the default feed, logged out, with no category or model filter applied. Logged-in, filtered or sorted views may rank differently. We did not test them.
- What the payload score is for. Unconfirmed. It does not order the feed we measured, and the leaderboard returned no data.
- Views. Zero on every card and not rendered. We cannot tell whether the counter is broken, disabled, or simply not exposed on cards.
PromptPlank publishes independently and has no affiliation with Faymas. No prompt titles or creator names from the sample are reproduced here.
The one-line version
Copies are a diary, not a dashboard. They record what people already wanted, on a feed that does not reward wanting it. That is not a reason to stop counting — it is a reason to stop refreshing. Publish, wait a day, write the number in your own notes, and start the next prompt. The next prompt is the only variable here you actually control.