dev.to's dashboard shows views per article, but not the question I actually had after 48 posts: which kinds of titles and tags get read? The public API has everything needed to answer it. Here is the script, and what it said about our own articles.

Median page views by title type for 48 of our dev.to articles

Disclosure: I build data tools for AI agents as Fast Data on Apify, and most of these 48 articles are about them. This post needs none of our tools; the script uses only the dev.to API and is in our public data-recipes repo.

Get an API key

dev.to → Settings → Extensions → "DEV Community API Keys" → generate. The key goes into an api-key header (not Authorization: Bearer). Store it as an environment variable, never in the script.

The script

import collections, datetime as dt, os, re, statistics import requests API = "https://dev.to/api/articles/me/published" headers = {"api-key": os.environ["DEVTO_API_KEY"], "User-Agent": "devto-stats/1.0"} articles, page = [], 1 while True: batch = requests.get(API, headers=headers, params={"per_page": 100, "page": page}, timeout=30).json() if not batch: break articles += batch page += 1 cutoff = (dt.datetime.now(dt.timezone.utc) - dt.timedelta(days=2)).isoformat() articles = [a for a in articles if a["published_at"] < cutoff] # give new posts 48 h first def kind(title): t = title.lower() if "n8n" in t: return "n8n recipe" if "how to" in t[:40]: return "how-to" if re.match(r"^(i|we) ", t): return "'I did X' story" if "?" in t: return "question title" return "other" def report(groups): rows = [(k, len(v), statistics.median(a["page_views_count"] for a in v), sum(a["public_reactions_count"] for a in v), sum(a["comments_count"] for a in v)) for k, v in groups.items() if len(v) >= 3] for k, n, med, react, comm in sorted(rows, key=lambda r: -r[2]): print(f"{k:22} n={n:3} median views={med:5} reactions={react:3} comments={comm:3}") by_kind, by_tag = collections.defaultdict(list), collections.defaultdict(list) for a in articles: by_kind[kind(a["title"])].append(a) for tag in a["tag_list"]: by_tag[tag].append(a) report(by_kind); print(); report(by_tag) 

Three things the API does that are easy to miss:

Detail Why it matters
/articles/me/published needs the key; /articles?username= does not only the authenticated endpoint returns page_views_count
pagination stops when a page comes back empty there is no total count in the response
tag_list is a list here, a comma-separated string on some other endpoints check the type before splitting

What it said about our 48 articles

Articles published 20 October to 13 November, measured 16 November, 1,320 views in total:

Title type Articles Median views Reactions Comments
How-to ("How to check a new supplier's company data … with Python") 6 35 1 1
"I did X" story ("I read all 308 open jobs at xAI …") 11 32 2 3
n8n recipe 14 30 7 4
other 9 20 7 1
Question title ("Which apprenticeships …?") 8 17 5 5

And by tag (only tags used at least 4 times; an article counts for each of its tags):

Tag Articles Median views
api 8 35
tutorial 5 34
jobs 6 31
python, ai, automation, n8n 10–23 each 30
ecommerce 12 25.5
webdev 8 20.5
fintech 6 11

What I take from it:

  • Question titles got about half the views. 17 vs 30–35. A title that promises a concrete answer or a concrete experience beats one that asks the reader to guess.
  • Reactions and comments are rare across the board: under one reaction per article on average. Views come from search and the feed; engagement comes from somewhere else.
  • Niche tags can cost reach. fintech had the lowest median (11). We now use it only as an extra tag, never as the main one.
  • More articles split the same attention. October articles have a median of 34 views, November articles 20, while we published more per day. Some of that is age, but the November articles are at least two days old too.

What the top articles in our tags do differently

For comparison, the API's ?tag=<tag>&top=7 endpoint returns the week's top articles per tag. In AI, MCP and automation the clear winners were essays with a contrarian thesis and long reading time: one 13-minute piece arguing that a popular protocol is "dead simple, and that's exactly the problem" had 182 reactions and 163 comments; another, about AI productivity claims vs what a QA team actually saw, had 110 comments. None of them were how-tos. If you want discussion, an argued opinion from experience seems to do what tutorials do not.

What this does not show

  • Small sample, one author. 48 articles from one account with a small following.
  • Title type is not the only difference. Topics, length and timing differ too; the categories are a regex, not a careful coding.
  • Views are not readers. dev.to counts page views; it does not tell you how many read to the end.

Try it

The full script: snippets/devto_stats.py. Change kind() to the title patterns you actually use.

Question: which title pattern works best for you on dev.to? I would love to compare with an account that writes opinion pieces rather than tutorials.