Original research2,119 sites16 months of Search Console data
Do AI SEO agents actually work? We tracked 2,119 sites to find out
Short answer: for most sites, yes, but slower and less evenly than the marketing suggests. The median site doubled its Google clicks about 7 months after it started. One in four ended up with fewer clicks. Here is all of it, including the parts that do not flatter us.
Published by the Outrank team · Data as of September 25, 2026
2.0×
Median clicks by month 7
vs the 28 days before joining, 596 sites
75.4%
Of sites gained clicks
1,383 sites, latest 28 days vs before joining
4.8×
Median impressions
27,296 → 131,458 per 28 days
24.4%
Of sites lost clicks
8.2% lost half or more
The headline in one chart
The median site went from 297 to 814 Google clicks a month
Clicks in the 28 days before joining Outrank against the latest 28 days, for 1,383 sites followed for a median of 218 days.
Median per site, per 28 days
Source: Google Search Console data connected to Outrank. Sites with at least 100 clicks across both windows and 90+ days since joining.
The curve
Three slow months, then it compounds
Every site is indexed to its own 28 days before joining (1×), then followed 28 days at a time. The line is the median; the shaded band is the middle half of sites.
28-day periods before and after joining Outrank
Source: 2,119 sites with at least 30 clicks in the 28 days before joining. Later periods include fewer sites because fewer have been with us that long.
+7%
First 28 days
60.5% of sites already above baseline
+27%
Days 57–84
New pages are indexed but mostly not on page one yet
2.0×
Days 197–224
78.2% of sites above baseline
The shape matters more than any single number. Most of the gain arrives after the third month, which fits how long new articles usually take to climb towards page one. Anyone who judges an SEO agent after four weeks is judging the flattest part of the curve.
The band widens over time. By month 7, the top quarter of sites were above 5.4× their baseline while the bottom quarter hovered around 1.1×. An AI SEO agent raises the odds, it does not flatten them.
Not everyone wins
One in four sites had fewer clicks
Change in 28-day clicks from before joining to the latest period, 1,383 sites.
113 sites (8%) ended with half their starting clicks or less. Hover a bar for details.
The largest single group landed between flat and 2×. Almost one in five grew tenfold or more, and three in four of those started under 100 clicks per 28 days. On the other side, 8.2% lost half their clicks or more. We would rather you hear that from us than find it in a Reddit thread.
By industry
It worked in every niche we could measure
Share of sites that gained clicks, and median 28-day clicks before and after, by the niche each site was classified into.
| Sports & fitness | 25 | 88% | 1,071 → 3,1062.9× |
|---|---|---|---|
| Automotive | 31 | 84% | 145 → 9856.8× |
| Fashion & beauty | 45 | 82% | 508 → 1,1542.3× |
| Home & garden | 118 | 82% | 282 → 1,1043.9× |
| Health & wellness | 95 | 81% | 490 → 1,4473.0× |
| Travel & hospitality | 32 | 81% | 136 → 9326.9× |
| B2B services | 111 | 79% | 124 → 3893.1× |
| Marketing & advertising | 113 | 78% | 154 → 5143.3× |
| HR & recruiting | 40 | 75% | 146 → 9856.7× |
| Finance & fintech | 38 | 74% | 214 → 4422.1× |
| Education & e-learning | 52 | 73% | 377 → 1,5674.2× |
| Real estate | 44 | 73% | 344 → 5861.7× |
| Legal & compliance | 50 | 72% | 737 → 1,5752.1× |
| E-commerce & retail | 55 | 71% | 875 → 2,2022.5× |
| Manufacturing & industrial | 28 | 71% | 528 → 8121.5× |
| SaaS & tech | 183 | 70% | 265 → 7652.9× |
Click a column to sort. Segments with fewer than 25 sites are left out.
No niche with at least 25 sites had fewer than 70% of sites growing. SaaS and tech, the most crowded category, sat at the bottom of the range; local and lifestyle niches like home and garden, fitness and automotive at the top.
Publishing
Sites that auto-publish grew more often
Grouped by how Outrank delivers articles to each site.
76.8%
Grew, with auto-publishing
1,247 sites · median 267 → 808 clicks
62.5%
Grew, without auto-publishing
136 sites · median 622 → 851 clicks
| Wix | 48 | 92% | 80 → 6618.3× |
|---|---|---|---|
| WordPress (REST) | 81 | 83% | 168 → 7674.6× |
| Shopify | 153 | 82% | 895 → 1,9932.2× |
| Webhook / custom | 332 | 78% | 210 → 9004.3× |
| WordPress plugin | 473 | 74% | 253 → 6732.7× |
| Notion | 26 | 73% | 806 → 2,0062.5× |
| Webflow | 67 | 72% | 368 → 8872.4× |
| Framer | 42 | 69% | 161 → 4592.9× |
| No auto-publishing | 136 | 63% | 622 → 8511.4× |
Click a column to sort. Segments with fewer than 25 sites are left out.
Sites without a publishing integration started with more traffic, which makes growth harder to show, so this is not a clean comparison. It is consistent with what we see anecdotally: the articles that get published on schedule are the ones that rank.
The case against
Four reasons to discount these numbers
Every study like this has weaknesses. These are ours.
Sites were already growing
In the 28 days before joining, the median site was up about 5.4% on the period before. If that trend had simply continued, it would reach about 1.5× by month 7. The observed 2.0× is above that, but not all of the growth is ours.
Self-selection
Only 4,689 sites have connected Search Console, and owners who bother to connect it may be more engaged than average.
Survivorship
Later periods only include sites still with us long enough to measure. Customers who left early after poor results drop out of the right side of the curve.
Correlation, not causation
There is no control group. Google updates, seasonality, product launches and other marketing all happen in the same months. The join date is when the site was added, not when the first article went live.
Behind the median
What it looks like for real businesses
Named customers, their own words, and their data, shared with permission.
Life Purpose App4.5×
Google clicks, May 2026 vs May 2025
Years of flat traffic, then 4.5× more clicks in a year
Martin Adams, Life Purpose App
24hourEDU1.2 → 30
Domain Rating on Ahrefs
A brand-new domain, from DR 1.2 to DR 30
Jesse Kennedy, 24hourEDU
Moonb22 → 43
Domain Rating on Ahrefs
From no SEO strategy to 2,213 AI citations
Lorenzo Nicolini, Moonb
The platform behind the data
What an AI SEO agent does at scale
1M+
Articles generated
1,002,810 as of September 25, 2026
473.7k
Verified backlinks
Checked live on the host page, as of September 4, 2026
10.9k
Sites in the link exchange
As of September 4, 2026
DR 26
Median linking site
Across the last 5,000 verified links
Methodology
How we ran the numbers
- Data: daily clicks and impressions from Google Search Console for the 4,689 sites that connected it to Outrank. Search Console keeps 16 months of history, which is the window we see.
- Join date: the date the site was first added to Outrank. Sites that joined between November 2024 and June 2026 have data on both sides of it. Duplicate entries for one domain use the earliest date.
- Curve: 2,119 sites with at least 30 clicks in the 28 days before joining. Each 28-day period is divided by that baseline; a period is included only if all its days are available.
- Before and after: 1,383 sites with a full 28-day window before joining, at least 90 days after, and 100+ clicks across both windows. We report medians, not averages, because a few very large sites would dominate an average.
- Privacy: no individual site is identified in the aggregate. The named case studies are published with each customer’s agreement.
Cite this study
Outrank (September 25, 2026). Do AI SEO agents work? Google Search Console data from 2,119 sites. https://www.outrank.so/case-studies/do-ai-seo-agents-work
Journalists, researchers and bloggers are welcome to quote these numbers and screenshot the charts with a link back to this page.
