CONSORT Flow Diagram Generator for Randomized Trials

A CONSORT flow diagram is the figure in a randomized trial report that follows every participant from eligibility screening to the analysis: how many were assessed, excluded and randomized, and for each arm how many received the intervention, stopped it, were lost and were analyzed. CONSORT stands for Consolidated Standards of Reporting Trials.

Item 22a of the CONSORT 2025 checklist asks, for each group, for "the numbers of participants who were randomly assigned, received intended intervention, and were analysed for the primary outcome". Item 22b asks for "losses and exclusions after randomisation, together with reasons". Most trial reports answer both in the flow diagram, which has four stages:

The template is spelled the British way (Enrolment, Randomised, Analysed). Write randomized and analyzed in your description and the boxes switch to US spelling; if your text mixes the two, check the headings.

CONSORT 2025 was published on 14 April 2025, at the same time in The BMJ, JAMA, The Lancet, Nature Medicine and PLOS Medicine (Hopewell et al.). It replaces CONSORT 2010 with a 30-item checklist, which adds seven items, revises three and deletes one, and it comes with an updated flow diagram. The four stages are the same; the follow-up and analysis boxes changed:

"For primary outcome" makes clear that the loss and analysis counts refer to the primary outcome, not to every measurement in the trial. Vizcept uses the 2025 wording by default. If your journal or protocol still uses the 2010 template, write "CONSORT 2010" in your description to get the 2010 wording and box order.

When we checked on 21 September 2026, three other CONSORT diagram generators (ConceptViz, SciDraw and ChatDiagram) still described their output as CONSORT 2010. The official 2025 Word template is on consort-spirit.org.

Reviewers trace each arm from randomization to analysis. Run these checks before you submit:

1. Assessed for eligibility − excluded = randomized.
2. The allocated counts of all arms add up to the number randomized.
3. In each arm, received + did not receive = allocated.
4. The reasons listed under a box add up to that box's total.
5. Analyzed is not simply allocated − lost − discontinued. People who stopped the intervention can still be followed up and analyzed, and an intention-to-treat analysis may include people whose outcome was imputed. Say which analysis population the box counts, intention-to-treat or per-protocol, in the caption or methods; CONSORT 2025 item 21b asks you to define who is included in each analysis.

Discontinued and lost to follow-up are separate counts, even though they share the follow-up box. Discontinued intervention means the participant stopped the treatment; lost to follow-up for primary outcome means you have no primary outcome data for them. One person can be counted in both.

A worked example, from the first figure in the examples below:

In this trial everyone who discontinued still gave primary outcome data, so analyzed = allocated − lost. That will not hold in every trial, which is why the generator is built not to fill a box by subtraction: the figure should show what happened, not what the arithmetic implies. The next section explains the check and what it cannot catch.

When an image model draws a flow diagram, every box and digit is painted as pixels from your description, and nothing ties a painted digit to the number you wrote. Vizcept splits the job in two:

• Reading: a text model (Anthropic Claude Sonnet 4.6, with OpenAI GPT-5.2 as a fallback) reads your paragraph and returns which count belongs in which box, as structured data.
• Drawing: code lays out the CONSORT template from that data and sets every word and digit in a font. No part of a typeset figure is painted by an image model; the cases the image model draws instead are listed at the end of this section.

Three rules follow:

1. Every count above zero must appear in your text. If the model returns a number that is not in your description (say it subtracted 197 from 200 to get 3), the number is removed and the box shows a blank "(n = )". The check looks for the number anywhere in your text, not next to that box, so a worked-out count that equals another number you wrote gets through, and so does 0, which the model is told to use only when you write none or all.
2. Missing reasons stay visible. A box with a count above zero but no reasons keeps "(give reasons)", which item 22b asks for.
3. The arithmetic is yours. The figure shows your numbers even when they do not add up; it does not balance them.

What still needs your eyes: whether each number landed in the right box, because that is the model's reading of your sentences, and the arm names and reasons, which the model copies from your text.

The image model draws the whole figure instead, so check every number on it, when:

• the design is outside the template: crossover trials, cluster trials that report clusters and participants separately, single-arm or non-randomized studies, and trials with more than five arms or more than four follow-up visits;
• your description contains a character the font cannot set, such as Chinese, Japanese, Korean, Arabic or Thai script, an emoji, or a symbol like ✓ or ☐;
• you attach an image, such as a sketch or an existing figure, before generating;
• reading, layout or saving fails partway, and the run is handed to the image model;
• you change a finished figure with Edit with AI, whether you describe the change, mark up the figure or recolor it.

The result does not yet show which way a figure was made. To change a count, edit your paragraph and generate again. More on how Vizcept draws each kind of figure.

Yes. CONSORT is the reporting guideline for randomized trials. Other study designs have their own guideline, and most of them ask for a flow diagram too:

This generator lays out randomized trials with two to five parallel arms. For a cohort study's flow, the AI diagram maker will draw it with an image model; check every number.

Yes, and you own the figure. Two things to do before you submit:

• Disclose the tool. A text model read your description to place the counts, so treat the figure as AI-assisted even though code drew it. Elsevier, Wiley and Taylor & Francis allow AI help with explanatory figures such as flow diagrams if you disclose it. Springer Nature does not permit "creating content from text prompts" but allows "diagrams derived from author-developed content", so ask the editor. The publisher policy table and disclosure statement generator covers ten publishers.
• Cite CONSORT 2025. Hopewell S, Chan AW, Collins GS, et al. CONSORT 2025 statement: updated guideline for reporting randomised trials. BMJ 2025;389:e081123.

Writing a thesis with a systematic review chapter as well? The PRISMA flow diagram generator draws that figure.

The options differ in what you type and who places each number.

If you want no AI anywhere near the figure, use the Word template or the R package. The R package is also the only option that works from participant-level data, so its counts cannot drift from your dataset. Prices and templates checked 21 September 2026.