Internet Is Fun 
It is a specific and maddening problem. You know you saw it. You can picture the page. You cannot remember a single word that would find it, and describing it to a search engine produces nothing but shopping results.
There is a decent order of operations for this, and it starts much closer to home than people usually try.
Start with your own history
Most of the time the answer is sitting in your browser. Ctrl+H (Cmd+Y on Mac) opens history, and history is searchable by text in the page title, not just the URL.
Two things people miss here. History syncs across devices if you are signed in, so the phone visit is on the laptop. And you can browse by date β if you can place roughly when you saw it, scrolling the right afternoon is often faster than any search.
Also check recently closed tabs (Ctrl+Shift+T repeatedly) if it was recent, and your downloads folder, which records the source site for each file.
Search your own messages
If anyone sent it to you, or you sent it to anyone, the link is in a chat log or an email. Searching your own outbox for “http” plus a remembered word is crude and works startlingly often.
The same applies to any notes app, any document where you paste things, and your own bookmarks β which are worth searching properly rather than scrolling, since most people have more than they think.
Search for the phrasing, not the subject
This is the main trick when history fails. Do not describe the site. Guess a phrase that would literally appear on it, and search for it in quotation marks.
Exact-phrase search behaves completely differently from descriptive search: it bypasses the semantic matching that keeps returning popular pages about your general topic. Combine it with a date restriction or a site restriction and you can get very specific. The full toolkit is in search operators that still work.
Try a different engine
Different indexes contain different things, and small sites in particular are indexed inconsistently. A query that fails on one engine sometimes works immediately on another, especially for older or low-traffic pages. See search engines that are not Google.
If you remember an image
Reverse image search works on anything you have a copy of, including a screenshot, and including a photograph of your own screen. If you saved the picture, or it is in your screenshots folder, this is often the shortest path.
If the site is gone
A remembered page that no longer exists is still findable. Web archives hold snapshots of enormous numbers of dead sites, and they are searchable by URL β so if you can reconstruct even part of the address, you can often get the page back. Archived link directories from the relevant era are also useful, because they list sites by category.
How web archives actually work covers the mechanics, and where to find the old web covers the older end specifically.
Last resort: ask people
There are communities dedicated to identifying half-remembered things, and they are remarkably good at it. A description containing the details you think are useless β the colour scheme, the animation, the odd thing in the corner β is exactly what solves these. The useless details are the identifying ones.
And then save it properly
Once you find it, do the thing that prevents the repeat. A bookmark in a folder you will not search is not saving it. A plain text file of links with a note about what each one is beats every bookmarking system, mostly because you will actually search a text file. More on that failure mode in digital hoarding.
