ended up being the most interesting slice by far — partly because they’re
the only category where people write real documentation.
Median description length across the 13,113 templates I have indexed is
2,238 characters. For comparison, the rest of my catalogue (models, MCP
servers, agents) has a median of 180. n8n template authors actually explain
what their workflow does, which made this worth measuring.
Sharing the numbers in case they’re useful to anyone here.
Which services show up most
Counts are how many of the 13,113 templates mention each term anywhere in
their title, description or tags:
google 7,770
sheets 5,471
slack 3,816
gmail 3,617
telegram 2,003
gemini 1,932
drive 1,892
claude 1,088
notion 1,049
airtable 1,011
whatsapp 769
calendar 766
discord 683
github 540
hubspot 523
supabase 414
wordpress 298
postgres 293
jira 268
shopify 241
stripe 163
xero 55
Methodology caveat so nobody is misled: these are text mentions, not
verified node usage. I tried measuring exact integration pairs and found my
own search was OR-matching tokens — “google sheets” was returning anything
containing “google” OR “sheets”, which inflated it to 7,835. So I threw out
every multi-word figure and kept only single tokens, which are reliable.
Treat these as “appears in” rather than “uses the node”.
The thing that surprised me
I sampled 500 templates and checked how many named two or more distinct
services in their title or tags.
57% named three or more.
That reframed how I think about this. The pattern isn’t “AI workflow” — it’s
glue. The model is one node in the middle of a chain that starts somewhere
boring and ends somewhere boring. Actual titles from the set:
Send engineering daily digest to Slack with Jira, GitHub, OpenAI, and Google Sheets
Triage law firm intake emails with Gmail, Slack, Google Sheets, and Claude
Score and route inbound leads with Claude, HubSpot, Slack, and Gmail
Answer Shopify support emails with Gmail, Shopify, Claude and Slack
Handle voice receptionist bookings with Vapi, Google Calendar and Sheets
None of those are “an AI workflow”. They’re a process that happens to call a
model partway through. Google Sheets outranking every vector database in the
set is the same story — people automate with the tools they already have open.
Other bits
Gemini (1,932) is ahead of Claude (1,088) in mentions, which I didn’t expect
given how the two get discussed.
Postgres is only 293. For a tool this popular with developers, durable
storage is overwhelmingly a spreadsheet rather than a database.
Stripe at 163 and Xero at 55 suggest money-touching automation is still rare
— plausibly because people are reluctant to let an automation move money.
Happy to run other cuts if anyone wants a specific breakdown. And if the
methodology looks wrong to you, say so — I’d rather fix the numbers than
defend them.
Full set is at AI Workflow Automations — Ready-to-Use Workflows | GenAiHub if useful.