SynthGap
AI-agent for novel meta-analysis topic discovery
Don't put your effort on a generic topic. Work on the best topic you can find.
Most review tools start once you already have a question. SynthGap starts before that. Topic discovery estimates where the potential is, from what has already been synthesized and which primary studies exist right now. Then deep analysis solidifies it: for the topics you choose, SynthGap refines each one, finds and reads the studies that belong to it, writes a critical report and shows which topics are actually the strongest.
Every other tool starts at step one. We start at step zero.
The usual review starts with a topic you already have. SynthGap starts one step earlier, by finding the best topic.
Every systematic review follows these steps
Most tools begin at step one. SynthGap adds the step before it.
-
Step 0 · Topic discovery Hard to find, and worth the most to get right A good topic is impactful, novel and feasible at the same time: a question that matters, that existing reviews have not answered, and that has enough primary studies to pool. Few topics are all three, and no one can read enough of the literature to tell which. The steps below are months of work, and all of it rides on this one choice.NovelFeasibleHigh impact
- PICO & protocolDefine the question and lock the plan
- Search strategyBuild and run the database searches
- Title & abstract screeningSift through every record found
- Full-text screeningRead each paper and decide
- Data extraction & risk of biasPull outcomes and appraise quality
- SynthesisPool results and grade the certainty
- Write-upPRISMA flow, checklist and manuscript
The best topics are right in front of us.
They are just hard to catch. There are too many papers and trials, and a topic only shows up when enough of them are read side by side.
A sliver of the literature
PubMed alone holds more than 30 million articles, and more arrive every day. However many you read, it is a sliver of them, and the best topics sit in the rest.
The whole field in view
Reading vast amounts of text is what an AI agent does well. It searches the whole literature, reads what it finds, checks every candidate against the reviews that already exist, and brings the best topics to the surface.
SynthGap does that reading, so you can work on the best topic you can find.
See how the best topics are found
Describe a clinical area. SynthGap searches the literature, checks every candidate against existing systematic reviews, and proposes topics worth running.
- Novel. Checked against existing reviews, so you see what is covered, outdated or missing.
- Valid. Built for methodological validity: each topic comes with a PICO framework and the real primary studies that make a sound review possible.
- High impact. Chosen to add something new to the literature and to answer an important question.
Explore the whole field, not just the top pick
The interactive map groups the literature into clusters and ranks them, so you can spot an active area that nobody has synthesized.
- Zoom and pan across the whole area you described.
- Click a cluster to see its studies and the systematic reviews that already cite them.
- Nothing hidden: topics the agent ruled out are shown next to the one it picked.
Then get the evidence behind each topic
Choose the topics you want and the agents do the groundwork for each one, side by side, ending in a report you can act on.
- Capture: every related study in the literature, searched until nothing new turns up.
- Selection: the final included studies, with the reason for every decision.
- Report: explores the full texts of all included studies, finds the strengths and weaknesses of each topic and writes a closing report ranking and comparing each topic.
Deep analysis comes with the Plus and Pro plans, one analysis for each topic you choose, charged from your monthly usage. Discovery is on every plan, the Free plan included.
An AI agent does the searching
You type a clinical area. The agent does the rest.
The agent searches the literature for high-impact, relevant primary studies. It then checks which of them are cited and what systematic reviews and meta-analyses have already been done, and comes up with the best topic, with its estimated PICO and the studies it estimates the topic would include.
Every topic comes with its evidence
Not just a suggestion — the framework, the studies, and the reasoning behind it.
Full PICO framework
Population, intervention, comparison, and outcome laid out for every proposed topic — not just a headline.
The primary studies, named
Every topic lists the actual studies — with PMIDs — that would go into the review, so you can verify before you commit.
Nothing hidden
Topics the agent considered and ruled out, and ones it flagged as still possible, are shown alongside the one it picked.
An interactive literature map
Explore the whole field yourself, clustered and ranked by novelty and recency, beyond just the top suggestion.
Pick the plan that fits how you work
Start free, no card needed. Every plan comes with usage that renews every month.
Free
For a first look
$0
No card needed. Usage resets every month.
Limited use
Sign up free- Discovery runs
- The interactive literature map
- Chat about your results
Plus
For finding and checking topics
$10/monthSave 20%
Billed $30 every 3 months. Usage resets every month.
Standard usage
Subscribe to PlusEverything in Free, plus
- Deep analysis: the topic refined, its studies, a critical report and a verdict
- Reserve a topic and draft a PROSPERO entry
- Top up when you run low
Pro
For scanning a whole field
$24/monthSave 20%
Billed $72 every 3 months. Usage resets every month.
More than 3x the usage of Plus
Subscribe to ProEverything in Plus, plus
- SweepsExplore a whole research area in one go: Agents do a broad survey of the topic and come up with most of the high impact available topics and every topic found comes back ranked. You choose which to analyse. A sweep takes 80 to 120 minutes, and chat stays open while it runs.
- A preview of what a meta-analysis on each topic would look likeEffect size calculationPooled ratios or mean differences, each with a confidence intervalHeterogeneity assessmentI², τ², Cochran's Q and a prediction intervalGraphsForest plot, funnel plot and more
Need more? You can top up by paying for credits, in any amount.
Prices are in US dollars. A plan renews automatically until you cancel, which you can do any time yourself from your account. See the Refund & Cancellation Policy.
Ready to find your next review?
Sign up free and describe a clinical area — your first discovery run starts in seconds.
Sign up free