OpenEvidence Search Helper: systematic reviews, GRADE, and Japanese input
Contents
Key features
・Prioritizes answers from synthesized literature such as comprehensive reviews (systematic reviews and meta-analyses: SR/MA)
・Includes GRADE-style recommendations
・Prevents the error of accidentally submitting halfway through Japanese IME conversion in the text area (a small but welcome fix)
For why I prioritize searching synthesized literature, please refer to my previous blog post.
https://note.com/embed/notes/ne2139f6ab3bc
Q: Is ChatGPT GPT-5 Thinking better for literature searches?
A: Yes, it is. Especially for research aimed at generating new insights in areas where evidence is not yet synthesized enough to have SRs/MAs, I think it is better to use GPT-5 Thinking.
For details on how to make the most of GPT-5 Thinking, see @genkAIjokyo's blog post.
https://note.com/embed/notes/n6426e5fbf51e
Q: Then what is the advantage of OpenEvidence?
A: Because it is free to use (whereas GPT-5 Thinking is paid), it is a viable option when you want to adopt a tool from the bottom up across an entire organization. In areas where knowledge such as SRs/MAs has already accumulated, it lets you quickly and easily check standard medical knowledge.
By way of analogy, GPT-5 Thinking is an optical telescope.
Even in sparse areas where evidence has not yet accumulated, it lives up to being the latest model: it excels at spotting distant stars, drawing constellations from just a few stars, and tracing lines of hypothesis across uncharted territory. On the other hand, a telescope lavishly equipped with high magnification and a large aperture takes some effort to set up, and the analysis takes a fair amount of time.
OpenEvidence, by contrast, is a pair of binoculars.
In areas where evidence has already accumulated, you can quickly zoom in on what is nearby and grasp the big picture (whereas things very far away will not come into focus no matter how hard you try—that is where GPT-5 Thinking comes in). In fields where a body of evidence has already been built up, even if an AI finds RCTs and lines up explanations for them, if a comprehensive review critically appraised and synthesized by humans already exists, checking that first is faster, safer, and more reliable. Starting with SRs/MAs reduces the cost of our own critical appraisal and lets us efficiently find answers to clinical questions during busy clinical hours.
Furthermore, because OpenEvidence is offered for free, I believe that used well, it can help narrow disparities in access to information.
A closer look at the features
What I built is simple, but I designed it to cut out both routine hassle and accidental submissions at once. First, it automatically appends an English prompt to the end of your search query asking for an explanation based on SRs/MAs and recommendations aligned with GRADE, guiding the tool to draw answers from the top of the 6S pyramid. It also fixes the issue where pressing Enter during Japanese IME conversion prematurely submits the query in the input field.
You can add the extension from this page.
Type "oe + space" in the address bar or click the icon to bring up the OpenEvidence search window. When you press the submit button, "Provide a comprehensive explanation drawing on systematic reviews and meta-analyses (SR/MA), and list recommendations according to the GRADE system." is automatically appended to your search query. (You can customize this appended text in the options.) As an aside, when using the smartphone app, I have registered the above prompt in my user dictionary under the shortcut "oe" (OpenEvidence). It is a simple feature, but please try comparing the results with a standard, unassisted search.