Editorial policy

Last updated 28 July 2026

This site publishes two kinds of thing: numbers produced by code, and guides written by people. Both can be wrong. This page describes how we try to keep them right and what happens when we fail.

Who writes this

Everything on CountEvery is written in-house by the small team at Taurus Alternative Strategies, LLC. There are no guest posts, no contributed articles and no syndicated content. We do not accept payment, products or links in exchange for coverage, and there are no affiliate links anywhere on the site. The only commercial relationship on any page is the Google advertising described in the privacy policy, and no advertiser sees an article before it is published or has any influence on what it says.

How the guides are researched

The reference material here falls into three categories, and we treat them differently:

• Things we can test. What Microsoft Word does with a hyphenated compound, how many characters a platform's field actually accepts, whether a counter splits on an em dash. We test these ourselves rather than repeating what other sites say, because what other sites say about this is frequently copied from a third site and out of date. Where a figure comes from our own testing, the page says so, and it says when we tested.
• Things with a published source. Reading rates, readability formulas, letter frequency tables and style guide rules all have primary literature behind them. We cite the primary source, not a summary of it. Where a widely repeated figure has no traceable source — the 200-words-per-minute reading rate is the classic example — we say that instead of repeating it.
• Things that are genuinely contested. Whether Y is a vowel, whether a hyphenated compound is one word, what counts as a sentence. We do not pretend these are settled. We describe the positions, say which one the tool implements, and explain why.

How numbers are labelled

Not every figure on this site is the same kind of fact, and mixing them up is how people end up trusting an estimate as if it were a measurement. Our convention:

• Counts — words, characters, sentences, letters, paragraphs — are exact, given the published rules. If they disagree with your expectation, either the rule or our implementation of it is at fault, and we want to hear about it.
• Syllables are algorithmic and approximate. English orthography does not allow an exact syllable count without a pronunciation dictionary.
• Reading time, speaking time and readability scores are models. They describe an average reader on ordinary prose and can be substantially wrong on a specific text.
• Conversion tables — words to pages, characters to SMS segments — depend on assumptions that we state alongside the number.

Corrections

When we get something wrong, we fix the page rather than quietly deleting it, and where the error was substantive we add a dated note at the foot of the page saying what changed. We do not annotate typo fixes or rewording.

If you find an error, tell us. Corrections from readers are the main way errors on a site like this get found, and reports that include the specific text that triggered the problem get fixed fastest. We aim to respond within two working days and to correct confirmed errors within a week.

How pages are kept current

Character limits and platform behaviour change without notice, which makes the limits reference the page most likely to go stale. It is reviewed quarterly, and every figure on it carries the date it was last verified. The counting rules on How we count change only when the implementation changes, and any change is noted there.

Every page shows an “updated” date in its byline. That date moves when the substance changes, not when the site is rebuilt, so a date that has not moved in six months means the page has been checked and left alone rather than forgotten.

On the use of AI

We use AI tools in drafting and editing, in the same way we use a spell checker and a reference manager: to produce first passes, to catch inconsistencies, and to argue with. No page is published without a person reading it end to end, checking the claims that can be checked, and taking responsibility for what it says. Where a figure came from a source, a human verified it against that source.

We mention this because a lot of tool sites are now assembled without that step, and the difference shows up exactly where it matters here: in figures that sound authoritative and are not true. If you find one of those on this site, it is our error to fix, not a tool's.