AI Edits Before and After a Website Launch

Two thirds of sites were never edited with AI after going live. Among 132 published We.Inc sites with full logs, 89 (67.4%) received no AI edit request after their first publish. Before launch, the median site took 3 edit requests, and the median site was published once.

Published sites by AI edit requests after first publish (n = 132)
Edit requests after launch	Sites	Share
None	89	67.4%
1 to 5	19	14.4%
6 or more	24	18.2%

Updated October 28, 2026

Launch day is often treated as the finish line. We wanted to know whether that holds for sites built with AI, where a change is one sentence away. So we lined up every logged AI request for a site against the moment it was first published.

Edit requests before launch

Published sites by AI edit requests before first publish (n = 132)
	Value
None	43 (32.6%)
1 to 3	35 (26.5%)
4 to 10	31 (23.5%)
11 or more	23 (17.4%)

Method and date range

Data: usage and deployment logs in our database, pulled by a read-only query on October 28, 2026. Projects were created between May 12, 2026, when per-request logging began, and October 27, 2026.

Sample: 132 projects with at least one deployment in ready state. An edit request is one distinct logged AI turn tied to the project. Launch is the first ready deployment.

As in our other studies, staff, internal, test and bot accounts were left out at query time. Bands with fewer than 20 sites were merged (6 to 20 and over 20 edits became one band of 6 or more).

Limitations

Sample of 132 is modest, and recent sites have had little time to collect edits after launch.
Manual edits in the code editor and visual edits that use no AI request are not logged here, so total changes are higher.
Some sites published before May 12 through older paths are not part of the sample.

Cite this study

We.Inc, "AI Edits Before and After a Website Launch", October 28, 2026, https://we.inc/research/edits-after-website-launch.

A split audience

The data shows two groups. A large group launches and leaves the site as it is. A smaller group treats the site as a living thing, with six or more requests after launch and several republishes.

Time to first publish was usually short, with a median of about 15 minutes from project creation, but the slowest quarter took 21 days or more. Fast launches and long tails both appear in the same product.

Different from our earlier timing figure

Our 2026 data hub reports about ten hours from signup to first publish. This study measures from project creation instead, and only for projects with deployment records from May 12 onward, which is why the numbers differ.

How it works

Selected projects created May 12 to October 27, 2026 (the period with per-request logs) that have at least one successful deployment, excluding team and test accounts.
Took the first successful deployment as the launch time.
Counted distinct AI requests logged for each project before and after that time.
Computed distributions and grouped sites into bands of at least 20.

Tips

Plan one review a month after launch. Most sites here were left untouched, which is how opening hours and prices go stale.
The busiest third of sites kept changing after launch; small, frequent fixes are normal and cheap with AI edits.
Publishing early is common. The median launch came about 15 minutes after the project was created.

Frequently asked questions

How many AI edits does a site need before launch?

The median was 3 requests and the middle half of sites used 0 to 6. About a third (43 of 132) published with no logged AI edit, often from a finished template.

Do people keep editing after they publish?

Most do not. 67.4% made no AI edit after launch, while 18.2% made six or more, so post-launch work is concentrated in a minority of sites.

How many times are sites republished?

The median published site had one successful deployment; a quarter had three or more.

Start for free or see all free tools. We.Inc is free to start, with no card needed. AI building is on paid plans from $20 a month.