Guide · how we choose what to build
What answers an AI cannot give
Last updated 2026-10-01
This page is the reasoning behind every tool on this site, written out so you can judge it rather than take it on trust. The short version: there is a dividing line between two kinds of question, and only one of them is worth building a page around.
It is also, incidentally, the constraint that decides which games and which features we work on next — so if you disagree with the reasoning here, you disagree with our roadmap, and that is a useful thing for you to know.
The dividing line
Some questions have one answer. "What does a Bond cost in OSRS?" has a number, and that number can be looked up, summarised, and delivered in a sentence. "How many codes are active in this game?" has a list. "Is Robux or V-Bucks cheaper per unit?" has a comparison that resolves once you have both prices.
All of those are what a summary can now answer directly, usually without the reader visiting anyone's page. That is not a complaint about summaries — it is an accurate description of what they are good at, and it means a page whose entire value is "here is the fact" has a shrinking job.
Now consider a different question:
"I have 400 million Cash, I earn about 30 million an hour, and I own Woodland Wings. What should I buy next, and should I keep going after that?"
This one has an answer too — a better-defined one than most, because it is arithmetic. But it cannot be answered without the reader's numbers. The answer is not a fact about the game; it is a fact about the game and this player's current position in it. Every player who asks gets a different answer, and the correct answer changes again next week when they have earned more.
That is the line. On one side: facts. On the other: answers that need your inputs.
Why the second kind holds up better
Three structural reasons, and none of them depend on guessing how search will change next year.
A summary can restate the answer, but it cannot replace the interaction. For a query like the one above, a competent summary can describe the method — work out cost per point of stat, compare it to what you have. It cannot know your Cash. So the useful output is either a method (which you then have to apply yourself) or a link to something that applies it. A calculator is the second thing, and it is the only version of the answer that is complete.
Interactive answers are re-consumed. “What is 1,000 Robux worth?” is asked once. “What is my upgrade worth?” is asked every time the player's Cash changes — which, in an idle game, is the entire play loop. The same page gets used repeatedly by the same person, with different numbers, and that is a fundamentally different relationship from a lookup that happens once and is never needed again.
A correct model survives changes that kill a fact. When a game is rebalanced, every page that quoted a tier price is wrong and every guide that recommended a strategy is stale. The model is not: the ladder's shape — price rising much faster than benefit — survives the rebalance, and so does the arithmetic. Only the inputs need updating, and if the page computes rather than quotes, that is a one-line change.
The test we apply before building anything: if the answer to the question never references a number the reader supplies, we do not build a page for it. It will keep being answered elsewhere, more cheaply than we can answer it, and building it would cost a reader's attention on a page that cannot be the best answer.
A worked distinction
Two questions about the same game, one on each side of the line.
"How much is 1,000 Robux in dollars?" This is a fact. Roblox publishes a bundle price; the division is one operation. A summary answers it in a sentence. There is no version of this page that is meaningfully better than that sentence, which is why we no longer build them — and, frankly, why building them was a mistake worth correcting.
"Should I keep grinding for Angelic Wings?" This one refuses to sit still. The answer depends on the player's current wing, their Cash, how fast they earn it, and what the zones they can already reach are paying them. There is no single sentence that is correct for more than one player, and the honest answer is arithmetic. That arithmetic is what the wings planner on this site runs.
The distinguishing feature is not complexity. The first question is easier, but that is exactly why it is the wrong thing to build a page around. Being slightly better than a summary at something a summary does well is not a position. Being the only thing that actually finishes the job is.
What this changes about a game economy site
Applied consistently, the dividing line reshapes an entire site, and being honest about that is part of the argument.
It means removal as well as addition. Pages whose only content is a restated fact — a converted amount, a ranked list of currencies, a comparison that resolves in a paragraph — stop earning their place, and the site is better without them. A smaller set of pages that each do something is worth more than a larger set that each repeat something, both to a reader and to a search engine trying to work out what a domain is for.
It means tools lead and prose follows. On this site the explanatory text on a tool page is generated from the same data the tool computes with, so the explanation cannot contradict the tool. Prose that is written separately from the thing it describes drifts within one revision.
And it means the honest answer to "what will people search for in three years" is that we are not trying to predict it. Interactive answers to questions with reader-supplied inputs are useful in any distribution channel the reader arrives through — search, a link, a message from a friend, a video description. A page whose value is a fact is only ever as valuable as the channel it was found in, and that channel is the part nobody controls.
Where we could still be wrong
Two real risks in this reasoning, stated plainly.
The first is that building tools is more expensive than writing pages, and the tools have to be good enough that a reader prefers them to a quick mental estimate. A calculator that adds friction — making you find four numbers before it tells you anything — loses to a summary that answers approximately in one line. The wings planner takes three inputs for exactly this reason, and takes sensible defaults for all of them so it produces an answer before you touch anything.
The second is the traffic pattern. Interactive tools historically convert a smaller share of visitors into sessions than text pages do, because the reader either needs the tool or does not. We are not going to pretend we have measured this for this site; we have not, and any figure we quoted would be someone else's data about a different site. What we can say is that the pages we removed were receiving effectively no traffic at all, so the comparison being replaced is not a bar that is hard to clear.
If you want the sourcing standard these tools hold themselves to — including where the numbers are weak and how that is labelled — that is the next guide.
More on how this site works: all guides · methodology · about