Under one percent. On 30 September 2026 the Faymas sitemap listed 11,813 individual prompt pages. We then reloaded the photography category ten times in a row — 200 card slots in total — and collected 107 distinct prompts. That is 0.9% of the catalogue, and it took ten full page loads to reach. No browsing surface on the site shows more than 20 prompts at a time, and not one of them publishes a total.
This guide measures how much of the Faymas catalogue a visitor can reach by browsing, using server-rendered page data only. No generation tests were run.
The index is two orders of magnitude bigger than any page
Start with the arithmetic, because it settles the question before any interpretation begins. Faymas publishes an XML sitemap. Reading it on 30 September 2026 returned 11,871 URLs in total. Here is how they split.
| Path | URLs | What it is |
|---|---|---|
| /prompt/ | 11,813 | One page per published prompt |
| /blog/ | 27 | Editorial posts |
| /prompts/ | 12 | Keyword collection pages |
| /category/ | 9 | The nine browse categories |
| /creator/resources/ | 4 | Creator guidelines |
| Other (home, about, terms…) | 6 | Static pages |
So the library is five figures. Every way of walking through it is two.
That gap is not a design flaw in the usual sense. It is a structural fact about how the site is built, and once you see it you stop browsing the way you used to.
Every surface caps at twenty
There are exactly three ways to move through the catalogue as a visitor: the home feed, the nine category pages, and search. All three render twenty cards. Not twenty-ish. Twenty.
| Surface | Cards per load | Date span in our snapshot | Highest copy count seen |
|---|---|---|---|
| Home feed | 20 | ~22 hours | 49 |
| Each category page | 20 | 19 Jul – 30 Sep 2026 | 1,911 |
| Search | 20, or fewer | not published | 163 (query "portrait") |
We read the card data straight out of each page rather than counting by eye, which let us pull creation timestamps and copy counts alongside the titles. That is where the interesting part starts.
The home feed is a one-day window
Every card in our homepage snapshot was created inside a 21.9-hour span: from 2026-09-29T12:35 UTC to 2026-09-30T10:29 UTC. Three consecutive fetches returned the identical twenty cards, so the feed is stable over minutes — but it is also strictly recent.
Which means the homepage is a new-arrivals board, not a best-of board. Nothing on it has had time to collect copies. The most-copied prompt on it carried 49.
Now compare. In the same hour, across the nine category pages, the most-copied prompt we found carried 1,911 copies. Thirty-nine times higher. The homepage cannot show you that prompt. It is too old for the window.
Category pages reshuffle on every single load
This is the finding that should change how you browse.
The home feed is stable. Category pages are not. We fetched /category/photography ten times back to back and recorded how many prompts in each response we had not seen before.
| Reload | New prompts | Cumulative unique |
|---|---|---|
| 1 | 20 | 20 |
| 2 | 0 | 20 |
| 3 | 17 | 37 |
| 4 | 17 | 54 |
| 5 | 15 | 69 |
| 6 | 13 | 82 |
| 7 | 13 | 95 |
| 8 | 12 | 107 |
| 9 | 0 | 107 |
| 10 | 0 | 107 |
Two hundred card slots. One hundred and seven prompts. By reload eight the well ran dry. We repeated the experiment on /category/wallpapers: eight reloads, 160 slots, 79 unique prompts, and the last reload produced nothing new.
Two explanations fit that shape, and we cannot separate them. Either the category genuinely holds around 110 items and serves a random 20 of them, or an edge cache is rotating a handful of pre-built variants. Both produce the same experience for you. We report both rather than pick one.
Either way the practical rule is identical: reloading is the only pagination a category page has. There are no page controls in the server-rendered HTML, no "load more", no total.
One detail worth noticing. Two of the nine categories — photography and social-media — returned only prompts from the last two days, exactly like the homepage. The other seven reached back into July and August. Photography is simply too high-volume for its random 20 to escape the last 48 hours. The slow categories are the ones with depth.
Paging through search mostly re-shows what you already saw
Search behaves differently again. Without a page parameter it is completely stable: four consecutive fetches of /search?q=portrait returned the same twenty prompts, in the same order.
Add &page=2 and you get a different twenty. Fetch page 2 twice and you get an identical set, so pages are deterministic. But they overlap heavily, and the overlap is the problem.
| Page | New prompts | Cumulative unique |
|---|---|---|
| 1 | 20 | 20 |
| 2 | 15 | 35 |
| 3 | 0 | 35 |
| 4 | 17 | 52 |
| 5 | 8 | 60 |
| 6 | 11 | 71 |
| 7 | 5 | 76 |
| 8 | 2 | 78 |
| 9 | 0 | 78 |
| 10 | 4 | 82 |
Ten pages. Two hundred slots. Eighty-two prompts. Fifty-nine percent of everything we were shown was something we had already been shown.
And no result total exists anywhere in the page data. There is no field for it. So there is no way to tell whether 82 is a tenth of the matches or all of them, and no way to know when to stop paging. You page until the novelty runs out, and the novelty runs out somewhere around page seven.
How to actually find things
- Never judge the catalogue from the homepage. It is a 22-hour window. Its ceiling in our snapshot was 49 copies against a catalogue that contains a 1,911-copy prompt.
- Use categories for depth, search for precision. Categories reach back to July and surface the heavily copied work. Search is where you go when you know the word.
- Reload instead of scrolling. On a category page, six to eight reloads is a full pass. One reload is 20% of one.
- Treat a short search result as complete. A query returning fewer than 20 has nothing hidden behind it — our "meme" query returned exactly one. Queries returning exactly 20 are the ones being truncated.
- Keep your own index. When the reachable set is under 1% of the indexed set and reshuffles between visits, a saved URL is the only stable reference you have. Copy the link the moment you see something worth keeping.
Point five is the one people resist. It feels redundant right up until the reload that does not bring the prompt back.
Where this measurement stops
- Sample size. The reload experiments ran on two of nine categories. The other seven may behave differently; we did not exhaust them.
- The 107 figure is a floor, not a pool size. Two readings fit it — a ~110-item category, or a rotating cache. We did not resolve which, and the distinction matters if you want to know how much you are missing.
- Search relevance is unmeasured. We counted results, not quality. Whether the 82 prompts we reached were the best 82 matches is a different question we did not test.
- Logged-in behaviour is untested. Every fetch was anonymous. A signed-in feed, favourites list or personalised ranking may expose more.
- No copy counts were verified against earnings. Faymas publishes no rate converting copies into money, so no financial claim is made here.
- Timestamps are snapshots. All figures come from 30 September 2026, roughly 11:30–11:50 UTC+8. The catalogue grows daily.
PromptPlank publishes independently and has no affiliation with Faymas. No prompt titles or creator names from these samples are reproduced here.
The browse budget
Think of it as a budget of twenty. Faymas will show you twenty things at a time, from a library of nearly twelve thousand, and it will reshuffle the twenty between visits. Spend the budget deliberately: go to categories when you want the good stuff, go to search when you want a specific thing, and reload rather than assuming you have seen the shelf.
The homepage is the one place not worth your twenty. It is a door, not a room.