This is just a rough draft. We should create (a) a shorter version based on our most common mistakes / omissions (b) different versions for different pages.
- Have all jargon words and abbreviations been replaced or explained?
- Have we cut out unnecessary hedging words?
Have all claims been given an example?
Can the summary be understood without reading the rest?
Is it easy to understand the main conclusions?
Have all cliches been removed?
Check all abstract words to see if they can be replaced by more concrete terms?
Adapted from GiveWell’s checklist
Does the bottom line seem correct, given the evidence that was provided?
Do any claims in this document seem overconfident?
Do any of the claims in the document need citations?
Does the document assert that X causes Y with effect size Z on the basis of low quality evidence, such as a single RCT?
Whenever someone makes a factual claim which is not common knowledge and there is a citation for that claim, include the citation.
Are there important remaining questions that the document does not clearly acknowledge?
Does the document defend any controversial claims with appropriate caution and qualification?
Does the document follow standard formatting?
Do all the links work?
Have footnotes copied and pasted relevant content to establish that what the body concludes on the basis of the footnote is true?
Was the research process clearly outlined somewhere?
Was the research process reasonable for answering the question? Were there any literature searches that should have been done but were not?
Did the document appropriately select and evaluate literature reviews on the question? 
Were important but difficult “judgment calls” clearly identified in the document. E.g., if there were two reasonable ways one could have done a certain calculation, was this fact noted and was there an explanation of why it was done in the particular way that it was? Or, if there were two reasonable ways that a literature search could have been limited, was that noted and was there an explanation of why it was done the way that it was?
Does the document carefully attend to questions of causal attribution and publication bias? See here for guidelines