162 AI Articles, Two Sites: What Google Actually Did With Them (2026 Data)
We published 162 AI-written articles on a cocktail site and an SEO blog. 70% got clicks on one, 23% on the other. Here is the Search Console data and what explains the gap.
Do AI-written articles rank on Google? We have a direct answer, because we publish them. Between August and November 2026, Bold Pilot wrote and published 162 articles on two sites: 100 on WhatDrink, a cocktail and spirits site, and 62 on this blog. Same pipeline, same keyword research, same writer. The results were very different.
On WhatDrink, 70 of 100 articles got clicks from Google and 55 reached an average position in the top 10. On this blog, 14 of 62 got clicks and 15 reached the top 10. Below is the Search Console data for every article, and what we think explains the gap.
The data in one table
All numbers come from Google Search Console, for each article's own URL, from the day it was published to 16 November 2026.
WhatDrink (cocktails) | Bold Pilot blog (SEO) | |
|---|---|---|
Articles published | 100 | 62 |
Got at least one impression | 85 (85%) | 55 (89%) |
Got at least one click | 70 (70%) | 14 (23%) |
Average position in the top 10 | 55 (55%) | 15 (24%) |
Total impressions | 331,791 | 41,081 |
Total clicks | 1,572 | 58 |
Median clicks per article | 2 | 0 |
Google indexed and showed almost all of them on both sites. Being indexed was not the problem. What differed was whether the article ranked high enough for anyone to click.
Finding 1: AI articles do rank, when the competition allows it
On WhatDrink, the best article ("Vodka Cocktails to Order at a Bar") earned 162 clicks at an average position of 6.3. Six articles earned 70 clicks or more, all at average positions between 5 and 8. None of them is unusual. They answer a specific question ("martini without vermouth", "Fireball alcohol content") better than the pages already ranking.
The whole site moved with them. In the first week of August, before the first article went up, WhatDrink got about 50 clicks a week from Google. In November it got 650 to 700 a week, and the Bold Pilot articles brought in about 300 of those, close to half.
So the answer to "can AI content rank?" is yes. Google's own guidance says it rewards helpful content however it is produced, and our data agrees: 55 AI-written articles reached page one on a site with modest authority.
Finding 2: the niche decided more than the writing
The same pipeline wrote both sites' articles. The difference is who they compete with.
A cocktail query like "negroni without vermouth" is answered mostly by recipe blogs and small food sites. An SEO query like "agency SEO software" is answered by Semrush, Ahrefs, HubSpot and dozens of SaaS companies with years of links behind them. Our "Agency SEO Software" article collected 15,926 impressions, more than any other article on this blog, at an average position of 63. Google kept testing it on page six and seven. It earned 10 clicks.
Writing quality cannot close a gap that big on its own. If your niche is crowded with established brands, expect AI articles (or any new articles) to sit on pages three to seven for months.
Finding 3: time matters, but less than you think
On WhatDrink, articles kept improving through their second month, then levelled off:
Days since publishing | Articles | Got clicks | Top 10 | Clicks |
|---|---|---|---|---|
0–29 | 30 | 17 | 12 | 123 |
30–59 | 30 | 28 | 23 | 767 |
60–89 | 31 | 20 | 17 | 469 |
90+ | 5 | 5 | 3 | 213 |
Articles in their second month did best. A month after publishing is a reasonable point to judge an article. Before that, most of what you see is Google still testing it.
Finding 4: longer did not mean better
Among articles at least 30 days old on WhatDrink:
Length | Articles | Top 10 | Median clicks |
|---|---|---|---|
1,500–2,499 words | 22 | 16 (73%) | 6 |
2,500–3,499 words | 34 | 21 (62%) | 8 |
3,500+ words | 9 | 5 (56%) | 5 |
The longest articles had the lowest share in the top 10. On this blog, the 26 articles over 3,500 words had a median of zero clicks. Length is not a ranking factor. Answering the question is.
Finding 5: a few articles do most of the work
On WhatDrink, the top 10% of articles earned 54% of all clicks. On this blog, the top 10% earned 88%. That is normal for search traffic, and it changes how to judge a content programme: do not judge it by the average article. Judge it by whether it produces a handful of winners, then improve those.
What this means if you publish AI articles
Pick keywords you can actually win. The niche and the competition on page one predicted results better than anything about the article itself.
Judge an article after 30 to 60 days, not after a week.
Do not chase length. 1,500 to 3,500 words was the sweet spot in our data.
Find your winners and improve them. Most clicks come from a small number of articles.
How we measured this
Every article in this study was written by Bold Pilot and published between 13 August and 16 November 2026. We matched each article's URL with its page-level data in Google Search Console (final data, not the fresh estimate), counting from its publish date to 16 November 2026. "Top 10" means an average position of 10 or better over that period. WhatDrink's owner agreed to share the site's numbers. We did not remove any article from the totals.
This is two sites, not two hundred. Treat the numbers as one honest data point, not a law.
✍️ Written by Ahmet, Founder of Bold Pilot
Ahmet builds Bold Pilot, writes the product and answers the support email. This blog runs on the same pipeline customers get, and he reviews what it publishes.
🔗 Bold Pilot is an AI-powered SEO automation platform that finds keywords a site is close…📢 Share this article
📚 More articles
A step-by-step Squarespace SEO checklist covering technical setup, on-page optimization, and built-in tools—so your site ranks faster. Here's how.
Find the keywords your site almost ranks for and turn them into traffic fast. Includes free tools, a prioritization method
Blog automation pro tools handle keyword research, writing, and publishing in one pipeline. See how they compare, what they cost
Learn how to automate blog posts from keyword research to publishing in one pipeline. Covers tools, workflows, and what to keep human. Here's how it works.
