Original data · updated monthly · updated

Brainbaby AI Index: how people create with AI

Anonymized, aggregate figures from Brainbaby, the AI workspace for chat, images, video, presentations and apps. Data through October 2026.

Updated

What is the Brainbaby AI Index?

The Brainbaby AI Index is a free monthly report of anonymized, aggregate data from Brainbaby’s AI platform. Through October 2026, at least 82.8% of saved AI images were photo edits, the median image prompt was 8 words and 66.1% of AI videos with a recorded mode started from an image. Figures are reusable under CC BY 4.0.

Publisher
Brainbaby (brainbaby.ai), from its own production database
Data through
October 2026 (figures read 21 November 2026)
Covers
14,286 saved AI images, 455 finished AI videos, 22,745 chat messages in three months
Privacy
Totals only; groups under 20 rows are merged into “other”; no prompts or personal data published
Updates
Monthly, on the 2nd, from read-only database queries
Reuse
Free under CC BY 4.0 with a link to this page

Headline findings

  • 82.8%or more of saved AI images were edits of an uploaded image
  • 8 wordsis the median AI image prompt
  • 7.6%of image prompts name a photorealistic style
  • 66.1%of AI videos started from an image, not text

How do people use AI image tools?

Edits, prompts and the styles people ask for, from every image saved to a Brainbaby image library.

  1. 82.8% or more of saved AI images were edits of an uploaded image

    At least 82.8% of the 14,286 AI images saved on Brainbaby were edits of an uploaded image; at least 11.6% were new images generated from a text prompt.

    Period: February 2026 to 21 November 2026. Method: Every image saved to a Brainbaby image library, classified by the model route that produced it. Routes that both edit and generate are counted in neither share.

    Source: Brainbaby AI Index, data through October 2026.

  2. 8 words is the median AI image prompt

    The median AI image prompt on Brainbaby is 8 words (46 characters), and 74.7% of prompts are 15 words or fewer.

    Period: February 2026 to 21 November 2026. Method: Words are runs of characters separated by spaces, counted over 14,279 saved images that have a prompt.

    Source: Brainbaby AI Index, data through October 2026.

  3. 7.6% of image prompts name a photorealistic style

    “Photorealistic” is the style named most often in Brainbaby image prompts: 7.6% of prompts ask for it, compared with 2.1% for cinematic and 0.7% for anime.

    Period: February 2026 to 21 November 2026. Method: Counts of prompts that contain one of a fixed list of style words (for example realistic, photorealistic, realista); a prompt can name more than one style. No other prompt content is read out.

    Source: Brainbaby AI Index, data through October 2026.

Length of AI image prompts on Brainbaby
GroupPromptsShare
1–5 words4,70933%
6–15 words5,94841.7%
16–40 words2,41216.9%
More than 40 words1,2108.5%
Image prompts that name a style (of 14,279 prompts; a prompt can name several)
GroupPromptsShare
Photorealistic1,0927.6%
Cinematic2952.1%
Anime1040.7%
3D render910.6%

How do people make AI videos?

Image-to-video against text-to-video, clip length and waiting time, from finished video generations.

  1. 66.1% of AI videos started from an image, not text

    66.1% of Brainbaby AI videos with a recorded mode started from an image rather than a text prompt (156 image-to-video and 80 text-to-video clips).

    Period: to 21 November 2026. Method: Finished videos whose mode was recorded; 219 older videos have no recorded mode and are left out of the share.

    Source: Brainbaby AI Index, data through October 2026.

  2. 98.9% of finished AI videos are 10 seconds or shorter

    98.9% of finished AI videos on Brainbaby are 10 seconds or shorter.

    Period: to 21 November 2026. Method: Length of every finished video with a recorded duration.

    Source: Brainbaby AI Index, data through October 2026.

  3. 2 min 37 s median wait for a text-to-video clip

    A text-to-video clip on Brainbaby took a median of 2 min 37 s from request to finished video (41 completed jobs).

    Period: to 21 November 2026. Method: Time between a video job being created and completed, for completed text-to-video jobs only. Small sample; read it as indicative.

    Source: Brainbaby AI Index, data through October 2026.

  4. 7 words median image-to-video prompt

    Image-to-video prompts are short: a median of 7 words, against 18 words for text-to-video prompts, because the image already supplies the scene.

    Period: to 21 November 2026. Method: Median word count of the prompts of finished videos, by recorded mode.

    Source: Brainbaby AI Index, data through October 2026.

What do people ask an AI assistant for?

What people send to Brainbaby chat and what kind of reply comes back.

  1. 23.6% of chat replies were images

    23.6% of Brainbaby chat replies were images rather than text (3,098 of 13,154 replies).

    Period: 1 August 2026 to 31 October 2026. Method: Assistant replies classified by the route that produced them (text answer, image, agent or web task, document or code, notice).

    Source: Brainbaby AI Index, data through October 2026.

  2. 8 words is the median chat message

    The median message people send to Brainbaby chat is 8 words (44 characters), but the mean is 416 characters because a few long pastes pull the average up.

    Period: 1 August 2026 to 31 October 2026. Method: 22,745 messages sent by people in complete calendar months.

    Source: Brainbaby AI Index, data through October 2026.

  3. 7,828 chat conversations started in October 2026

    7,828 chat conversations were started on Brainbaby in October 2026, 81% more than in September 2026 (4,326).

    Period: September 2026 and October 2026. Method: Conversations still stored, by the calendar month (UTC) they were started.

    Source: Brainbaby AI Index, data through October 2026.

Brainbaby chat replies by type, 1 August 2026 to 31 October 2026
GroupRepliesShare
Text answer8,61665.5%
Image3,09823.6%
Notice or routing message1,2549.5%
Agent or web task1551.2%
Document or code310.2%

Who uses AI creation tools, and on which devices?

Browser language, country and device of accounts, from sign-in records.

  1. 36.1% of accounts use Portuguese as their browser language

    Portuguese is the most common browser language among Brainbaby accounts (36.1%), ahead of English (20.2%) and Spanish (19%).

    Period: 19 November 2026 to 21 November 2026. Method: The browser language of each account’s latest sign-in, over 1,355 accounts that signed in during the period. Languages with fewer than 20 accounts are grouped as other.

    Source: Brainbaby AI Index, data through October 2026.

  2. 35.4% of accounts are in Brazil

    Brazil is the largest country among Brainbaby accounts by time zone (35.4%), followed by the United States (6.3%) and Mexico (6%).

    Period: 19 November 2026 to 21 November 2026. Method: Country inferred from the time zone the browser reports at sign-in, only where a time zone belongs to a single country; other zones and countries with fewer than 20 accounts are grouped as other.

    Source: Brainbaby AI Index, data through October 2026.

  3. 70.7% of accounts signed in from Android

    70.7% of Brainbaby accounts last signed in from Android; 15.4% used iOS and 11.4% Windows.

    Period: 19 November 2026 to 21 November 2026. Method: The platform the browser or app reports at sign-in, reduced to the operating system.

    Source: Brainbaby AI Index, data through October 2026.

Accounts by browser language
GroupAccountsShare
Portuguese48936.1%
English27420.2%
Spanish25819%
Arabic14510.7%
Indonesian574.2%
Turkish503.7%
German292.1%
Other (groups under the minimum size, or unattributed)533.9%
Accounts by country (from time zone)
GroupAccountsShare
Brazil48035.4%
United States856.3%
Mexico816%
Indonesia644.7%
Türkiye564.1%
Egypt513.8%
Spain463.4%
Iraq342.5%
Germany332.4%
United Kingdom302.2%
Argentina272%
Saudi Arabia261.9%
India251.8%
Other (groups under the minimum size, or unattributed)31723.4%
Accounts by device platform
GroupAccountsShare
Android95870.7%
iOS20815.4%
Windows15511.4%
Other (groups under the minimum size, or unattributed)342.5%

How long do AI presentations and app builds take?

Deck length and build times from completed jobs.

  1. 11 slides in the median AI presentation

    The median AI presentation made on Brainbaby has 11 slides (128 completed decks).

    Period: to 21 November 2026. Method: Slide count of completed presentations.

    Source: Brainbaby AI Index, data through October 2026.

  2. 63 seconds median AI web app build

    A completed AI web app build on Brainbaby took a median of 63 seconds, and a mobile app build 2 min 10 s (42 and 75 completed builds).

    Period: to 21 November 2026. Method: Time from a build job starting to finishing, for completed builds only.

    Source: Brainbaby AI Index, data through October 2026.

How is AI use on Brainbaby changing month by month?

Chat conversations started per month on Brainbaby05k10kJanuary 2026: 377January 2026: 377Jan 26February 2026: 1,900February 2026: 1,900Feb 26March 2026: 4,944March 2026: 4,944Mar 26April 2026: 2,713April 2026: 2,713Apr 26May 2026: 1,903May 2026: 1,903May 26June 2026: 2,407June 2026: 2,407Jun 26July 2026: 2,501July 2026: 2,501Jul 26August 2026: 2,519August 2026: 2,519Aug 26September 2026: 4,326September 2026: 4,326Sep 26October 2026: 7,828October 2026: 7,828Oct 26
Chat conversations started per calendar month (UTC). Source: Brainbaby AI Index, updated 21 November 2026.
AI images saved per month on Brainbaby02.5k5kJanuary 2026: 0Jan 26February 2026: 34February 2026: 34Feb 26March 2026: 2,108March 2026: 2,108Mar 26April 2026: 1,410April 2026: 1,410Apr 26May 2026: 1,213May 2026: 1,213May 26June 2026: 3,392June 2026: 3,392Jun 26July 2026: 1,508July 2026: 1,508Jul 26August 2026: 784August 2026: 784Aug 26September 2026: 695September 2026: 695Sep 26October 2026: 1,291October 2026: 1,291Oct 26
AI images saved to Brainbaby image libraries per calendar month. More detail on images and videos: AI image & video generation statistics.
Monthly figures (complete months)
MonthImagesof which editsVideosChat messagesConversations
January 2026007543377
February 202634073,9681,900
March 20262,1081,387315,1984,944
April 20261,4101,0751334,6882,713
May 20261,2131,0681233,1801,903
June 20263,3923,077853,8882,407
July 20261,5081,394183,5992,501
August 2026784747126,2412,519
September 202669550895,5154,326
October 20261,291984910,9897,828

How was this data collected?

  • All figures come from read-only, aggregate database queries run by a scheduled job. Only counts, shares, medians and means leave the database; no prompt text, file, account id, email or IP address is read into the report.
  • Breakdowns drop every group with fewer than 20 rows into “other”, and a share is published only when it rests on at least 100 rows.
  • Months are calendar months in UTC. Monthly series stop at the latest complete month; all-time totals run to the day the figures were read.
  • Images are those saved to a Brainbaby image library. Edits and text-to-image generations are told apart by the model route that produced them; routes that do both are counted in neither share, so both shares are lower bounds.
  • Videos are generations that finished successfully. Image-to-video and text-to-video shares use only the videos whose mode was recorded.
  • Account language, country and platform come from each account’s latest sign-in record during the stated period. Country is inferred from the browser’s time zone only where a zone belongs to one country.
  • Chat reply types are assigned from the internal route that produced each reply, not from reading the reply.

These are Brainbaby’s own figures, not estimates for the whole AI market. They reflect Brainbaby’s audience, which is weighted towards Portuguese- and Spanish-speaking users and towards image editing, and product changes during the period.

First published . Last updated .

Can I download the data?

Yes. The same figures are published as open data under the Creative Commons Attribution 4.0 licence.

Cite this report

Using a figure? Quote the sentence under it and link to this page or to the finding’s anchor so readers can check the method and the latest month.

Brainbaby. (2026). Brainbaby AI Index: how people create with AI (data through October 2026). https://brainbaby.ai/reports/ai-creation-index
<a href="https://brainbaby.ai/reports/ai-creation-index">Brainbaby AI Index</a> (data through October 2026)

Questions about the AI Index

What is the Brainbaby AI Index?

The Brainbaby AI Index is a free monthly report of anonymized, aggregate data from Brainbaby’s AI platform. Through October 2026, at least 82.8% of saved AI images were photo edits, the median image prompt was 8 words and 66.1% of AI videos with a recorded mode started from an image. Figures are reusable under CC BY 4.0.

Do people use AI image tools more to edit photos or to generate new images?

On Brainbaby, mostly to edit. 82.8% of the 14,286 AI images saved were edits of an uploaded image, and at least 11.6% were generated from a text prompt. The rest came from routes that do both.

How long is a typical AI image prompt?

The median AI image prompt on Brainbaby is 8 words (46 characters). Most prompts are short instructions rather than long descriptions.

How is the AI Index data collected and anonymized?

A scheduled job runs read-only aggregate queries against Brainbaby’s production database once a month. Only totals, shares and medians are kept, groups with fewer than 20 rows are merged into “other”, and no prompt text or personal data is published.

How often is the Brainbaby AI Index updated?

Monthly, on the 2nd of each month, with the previous month’s figures. This edition has data through October 2026 and was updated on 21 November 2026.

Can I use the Brainbaby AI Index figures in my article or research?

Yes. The figures and the downloadable dataset are free to reuse under the Creative Commons Attribution 4.0 licence. Credit Brainbaby AI Index and link to https://brainbaby.ai/reports/ai-creation-index.

Related