Most of your traffic isn't people.
Here is a year of one online store, X-rayed: who visits, how they arrive, and how much of it your analytics counts as customers.
58%
of 4.8M sessions last month were likely automated
21%
came through residential proxies, the hardest kind to see
6%
were likely AI agents, three times the share a year ago
Sample store. Illustrative figures, not customer data.
People went from 59% to 42%.
The store did not lose customers. Automation grew around them. Residential proxies grew fastest, because they are cheap and most bot tools trust a home address.
NovDecJanFebMarAprMayJunJulAugSepOct
Likely peopleResidential proxiesScrapers and botsDatacenter and VPNAI agents
Where it comes in.
Automated share by how the visit arrived. Scrapers open product pages directly. Paid clicks are cleaner, but that is the traffic you pay for one click at a time.
X-ray your ad spend →Direct
2M sessions
Product pages opened with no outside referrer
64%
Social
140K sessions
Mostly preview fetchers and click farms
24%
Email and affiliate
96K sessions
Link scanners open every link in every email
22%
Google Shopping
720K sessions
Price monitors arriving through listing links
12%
AI chat referrals
61K sessions
People who clicked a link inside an AI answer
11%
Paid search
560K sessions
The share you pay for, click by click
10%
Microsoft Ads
85K sessions
Smaller, and cleaner
7%
Organic search
910K sessions
Including crawler impostors that claim to be Googlebot
6%
Datacenter bots cluster where the clouds are. Proxies are everywhere your customers are.
Virginia, Oregon, Ohio and Iowa host the big cloud regions. Everywhere else, the automated share is residential proxies, which look like the people who live there.
AK
ME
VT
NH
WA
ID
MT
ND
MN
IL
WI
MI
NY
MA
RI
OR
NV
WY
SD
IA
IN
OH
PA
NJ
CT
CA
UT
CO
NE
MO
KY
WV
VA
MD
DE
AZ
NM
KS
AR
TN
NC
SC
DC
OK
LA
MS
AL
GA
HI
TX
FL
Fewer botsMore bots
1.8×
Growth in residential-proxy sessions over the year
6 in 10
Product page views with no outside referrer come from automation
11 s
Median time on a product page for a scraper on a home address. People: 1 min 48 s
How we count.
- What we count
- Sessions: one visitor, one site, up to 30 minutes apart. Each session gets one class, the strongest one its evidence supports.
- Likely people
- Sessions with no automation evidence. Anything we are not sure about stays here, so the automated share is a floor, and no visit is ever called a person for certain.
- Residential proxies
- Visits relayed through someone's home internet connection, identified from connection evidence on the visit and across visits.
- AI agents
- Agents that name themselves, agents verified against their operator's network, and automated browsers flagged by their behaviour.
- This page
- Every figure here is for one illustrative store. Your X-ray uses your own traffic and updates every hour.