In one 90-day observation of a single Securify staging store (19 human sessions, possibly including test traffic), 320 of 339 sessions were classified as bots. The completed-checkout rate was 0.59% across all sessions and 10.53% across human sessions only.
When a report includes bots and shoppers in one denominator, merchants can mistake traffic composition for a conversion problem.
320 / 339
sessions classified as bots in a direct 90-day ShopifyQL query on one Securify staging store, run 2026-11-02
0.59%
completed-checkout rate using all 339 sessions: 2 completed checkouts (same staging store and query)
10.53%
human-only calculation using 19 human sessions: 2 completed checkouts (same staging store and query)
By default
Shopify filters identified bot sessions out of session-related reports once its session-measurement update reaches a store (rollout Sept 21–23, 2026)
Source: Shopify Help Center: bot filteringEvidence reviewed
This merchant-first evidence page explains how bot-heavy and AI-referred traffic can change a Shopify dashboard before a merchant changes ads, storefront design, checkout, or fraud controls. Securify is developing a read-only Data Integrity Analytics view; it does not rewrite Shopify analytics or automatically block visitors.
Offer name: Run my Data Integrity Report
A direct ShopifyQL query was executed on a Securify staging store at 2026-11-02T06:10:31Z for the preceding 90 days, deliberately grouped by human_or_bot_session and agentic_referring_channel to include both bot and human rows. The all-session rate is 2 divided by 339; the human-only calculation is 2 divided by 19. A separate Ranksy attribution dataset is supporting evidence and is not joined to these storefront sessions.
This is one staging store with only 19 human sessions, and the window can include setup or test-generated traffic that was not separately labeled, so it is not a Shopify benchmark. Shopify filters identified bots from supported session reports by default; this query intentionally included both bot and human rows. Shopify's classifications are used as reported and were not independently validated. The result does not prove that ChatGPT caused a purchase or that Securify caused, blocked, or corrected any session. Automation is not the same as fraud, and uncertain sessions must remain unknown.
Earlier control. Clearer signals. Less downstream mess.
The proposed read-only view would keep Shopify's raw session totals visible and unchanged.
It would add a separate commerce-adjusted view with a traffic-class breakdown.
It would explain each classification and preserve an evidence trail, including unknowns.
Conversion divides outcomes by sessions. In this observation, adding 320 bot sessions to the denominator changed the completed-checkout calculation from 10.53% for human sessions to 0.59% for all sessions. It did not change the two completed checkouts.
Not in supported session reports. Shopify filters identified bot sessions out of those reports by default. Merchants can change the report filter to include bots or add the human-or-bot dimension. This observation intentionally included both classifications.
Yes. Shopify reported one human session with ChatGPT as the agentic referring channel. It added to cart and did not complete checkout. One session does not establish repeatability or commercial value.
In a separate Securify-internal dataset (from Ranksy, the app-install tracking tool Securify uses; not independently published), Securify recorded 29 AI-assistant sessions, 5 installs, 1 trial start, 1 trial conversion, and $5.99 in attributed revenue. Ranksy does not identify ChatGPT specifically, the datasets are not joined, and the result does not prove causality, repeatability, or scale.
Free on every plan · no credit card · about 2 minutes
Guide
Broad guide for merchants trying to understand how bad traffic turns into fraud cost, manual work, and noisy data.
Read guideBots and VPNs
Operational guide for merchants seeing suspicious visitors, VPN traffic, scraping, and blocking failures.
Read guidePeak season
How BFCM, campaigns, and promo spikes compress response time and make bad traffic more expensive.
Read guide