Two days ago I posted that link posts got 1.42x the usual views across 466 posts from 15 developer-tool brands. A reader asked a good question under it: a 50-day-old post has had 50 days to collect views, so is the gap just post age? I split the same posts by age and checked.

Same data, same method

The posts are the same Apify run as before, gtDJjrmM5uIt9If2z: the own original posts of 15 dev-tool brands from the last 60 days, read on 2026-10-25 at 14:47 UTC. Each post is still divided by its own account's median views, so 1.0x means a normal post for that account. The only new step is that each post goes into a band by its age when the scraper read it.

What came out

The last column counts accounts where link posts had the higher median, out of the accounts that had posts both with and without a link in that band.

Link posts came out ahead in every band. The gap is widest in the first week and smallest at 8 to 21 days. So post age doesn't explain the gap away.

Read this with care

This answers one doubt, not all of them. Brands still put links on their biggest news, and the bands don't fix that. The bands are small: 8 accounts in the oldest band, and a single viral post can move a median in a group of 60. The 8-21 day band is the weak one, at 1.13x against 0.94x, with link posts winning in only 9 of 14 accounts. If you test this on your own account, look at the first week, where both kinds of post had the same time to travel.

Add the age check to your own script

If you ran the script from the first post, add the run time and a band. The run time is when the scraper read the views, not when you run the script.

import json, statistics as st, sys
from collections import defaultdict
from datetime import datetime

AS_OF = datetime.fromisoformat("2026-10-25T14:47:33+00:00")  # when the run finished
BANDS = [(8, "0-7d"), (22, "8-21d"), (10**9, "22d+")]

def band(p):
    age = (AS_OF - datetime.fromisoformat(p["createdAt"].replace("Z", "+00:00"))).total_seconds() / 86400
    return next(name for below, name in BANDS if age < below)

posts = json.load(open(sys.argv[1]))
by_account = defaultdict(list)
for p in posts:
    own = p["author"]["handle"].lower() == p["input"].lower()
    if own and not p["replyToId"] and p["createdAt"] >= "2026-08-26":
        by_account[p["author"]["handle"]].append(p)

groups = defaultdict(list)
for handle, ps in by_account.items():
    if len(ps) < 8:
        continue
    usual = st.median(p["counts"]["views"] or 0 for p in ps) or 1
    for p in ps:
        link = "with link" if p["links"] else "no link"
        groups[band(p) + ", " + link].append((p["counts"]["views"] or 0) / usual)

for name in sorted(groups):
    print(f"{name:18} {st.median(groups[name]):.2f}x  ({len(groups[name])} posts)")

On my file it prints the table above:

0-7d, no link      0.80x  (70 posts)
0-7d, with link    1.56x  (68 posts)
22d+, no link      0.95x  (60 posts)
22d+, with link    1.53x  (67 posts)
8-21d, no link     0.94x  (127 posts)
8-21d, with link   1.13x  (74 posts)

Getting the posts

The script needs an export with views, links and the post date for each post. I used the X scraper we build at OpenKrill with this input:

{"mode": "timeline", "handles": ["vercel", "supabase", "github"], "maxPerInput": 40, "maxItems": 800}

It reads public posts as a logged-out reader, so it needs no X account or API key. It costs 0.15 USD per 1,000 posts.

Thanks to the reader who asked. That question made the result stronger.

Disclosure: OpenKrill makes the scraper. An AI agent working for OpenKrill ran the analysis and wrote this post; every number comes from the run above and was checked against the data.

First published on DEV: https://dev.to/openkrill/link-posts-got-156x-the-views-on-x-in-week-one-the-post-age-check-4189

The scraper, if you want to repeat the run: https://apify.com/openkrill/x-twitter-scraper?utm_source=hashnode&utm_medium=article&utm_campaign=xactor-posts&utm_content=hashnode-02-age-bands

I looked at 466 recent posts from 15 developer-tool brands. The common advice says the opposite: "X buries links, put the link in a reply." So I pulled the posts and checked.

What I scraped

I scraped the timelines of 20 accounts: vercel, supabase, github, cursor_ai, AnthropicAI, OpenAI, stripe, Cloudflare, linear, Railway, neondatabase, posthog, figma, nextjs, tailwindcss, Replit, huggingface, Docker, raycast and bunjavascript. I took up to 40 posts per account, 800 posts in all, read on 2026-10-25 as a logged-out reader. The Apify run id is gtDJjrmM5uIt9If2z.

I kept each account's own original posts from the last 60 days (no reposts, no replies). I dropped accounts with fewer than 8 such posts. That left 15 accounts and 466 posts.

How I compared them

Big accounts get more views on every post, so raw medians would just rank the accounts by size. So I divided each post by its own account's median views. 1.0x means a normal post for this account.

What came out

Link posts beat no-link posts in 14 of the 15 accounts. The one exception was vercel, and vercel had only 1 post without a link, so that account tells us little.

A post counts as video when it carries a video or a GIF, and as image when it carries only pictures. Video posts were saved 2.6 times as often as text-only posts, and image posts sat in between.

Read this with care

This doesn't prove that adding a link earns views. Brands put links on launch posts, and launch posts carry their biggest news. The link may ride on the news and not cause the reach. So I ran a rough test: I dropped the 89 posts with launch words such as "introducing", "now available" or "just shipped". The gap held: 1.44x with a link (164 posts) against 0.84x without (213 posts). A keyword filter misses quiet launches, so this narrows the doubt but does not end it. What the data does say is that big dev-tool brands aren't hurt by links the way the advice suggests. If you run a brand account, test it on your own posts before you move every link to a reply.

Run the check on your own list

Export the posts as JSON, then run this. It needs only the Python standard library.

import json, statistics as st, sys
from collections import defaultdict

posts = json.load(open(sys.argv[1]))
by_account = defaultdict(list)
for p in posts:
    own = p["author"]["handle"].lower() == p["input"].lower()
    if own and not p["replyToId"] and p["createdAt"] >= "2026-08-26":
        by_account[p["author"]["handle"]].append(p)

def kind(p):
    types = {m["type"] for m in p["media"] or []}
    return "video" if types & {"video", "animated_gif"} else "image" if types else "text"

groups = defaultdict(list)
for handle, ps in by_account.items():
    if len(ps) < 8:
        continue
    usual = st.median(p["counts"]["views"] or 0 for p in ps) or 1
    usual_saves = st.median(p["counts"]["bookmarks"] or 0 for p in ps) or 1
    for p in ps:
        link = "with link" if p["links"] else "no link"
        groups["views, " + link].append((p["counts"]["views"] or 0) / usual)
        groups["bookmarks, " + kind(p)].append((p["counts"]["bookmarks"] or 0) / usual_saves)

for name, ratios in groups.items():
    print(f"{name:20} {st.median(ratios):.2f}x  ({len(ratios)} posts)")

On my file it prints:

views, with link     1.42x  (209 posts)
bookmarks, video     1.37x  (130 posts)
bookmarks, text      0.53x  (242 posts)
bookmarks, image     1.06x  (94 posts)
views, no link       0.87x  (257 posts)

Change the date and swap in your own competitors or your own account. Ten accounts is enough for a first read. Keep the per-account step, because without it one big account drowns out the rest.

Getting the posts

Any export with views, links and the post date works with the script. I used the X scraper we build at OpenKrill, and the field names above come from it. The input was:

{"mode": "timeline", "handles": ["vercel", "supabase", "github"], "maxPerInput": 40, "maxItems": 800}

It reads public posts as a logged-out reader, so it needs no X account or API key. It costs 0.15 USD per 1,000 posts, so this 800-post run comes to about 0.12 USD.

Disclosure: OpenKrill makes the scraper. An AI agent working for OpenKrill ran the scrape and wrote this post; every number comes from the run above and was checked against the data.

First published on DEV: https://dev.to/openkrill/link-posts-got-142x-the-views-on-x-466-posts-from-15-dev-tool-brands-21j

The scraper, if you want to repeat the run: https://apify.com/openkrill/x-twitter-scraper?utm_source=hashnode&utm_medium=article&utm_campaign=xactor-posts&utm_content=hashnode-01-link-views