Checklist
How many ad requests go out before the banner reads GPC? Four stores tested
We loaded 4 Shopify stores with GPC on and clicked nothing. On Store O, two runs sent 62 and 63 ad requests before its banner recorded GPC.
In brief
Tested 2026-10-27 (UTC+8). Four Shopify stores, three ways of handling GPC: a OneTrust cookie, an Osano consent object, and the Shopify Customer Privacy API. On Store O, OneTrust recorded GPC at 4069 ms and 4336 ms. Before that marker the two runs sent 62 and 63 ad requests (40 and 41 excluding script downloads), from 13 vendors. The Osano stores returned an opt-out marker no later than 1004 ms, with at most 2 script downloads before it. Store P has no CMP. It sent 5 and 8 ad requests before Shopify first read sale_of_data as "no". Each run used a new browser profile, a direct US visit, Sec-GPC plus the JavaScript flag, and no clicks. Two GPC runs per store. The paired protocol is the Shopify GPC honor test. Not legal advice.
When did each banner process GPC?
Tested 2026-10-27 (UTC+8), from a direct US connection in Virginia. The browser was headless Chrome, with a new profile each run, English, and the America/Los_Angeles timezone. GPC runs sent a Sec-GPC: 1 header and set navigator.globalPrivacyControl to true before any page script ran. We opened the homepage, clicked nothing, and waited 20 seconds after DOMContentLoaded. Each store has two GPC runs, listed separately, and one run without GPC.
The marker is the first sign that the store processed GPC. On Store O it is the moment OneTrust wrote the OptanonConsent cookie with isGpcEnabled=1. That write time is exact. On Store K and Store R it is the first time Osano.cm.getConsent() returned OPT_OUT: ACCEPT. Those stores set no consent cookie with a value (Osano only cleared old cookies), so the cookie is not the marker. The read is every 100 ms, so the time is an upper bound: no later than the millisecond in the table. On Store P there is no CMP. The marker is the first time Shopify.customerPrivacy.currentVisitorConsent() returned sale_of_data: "no". That read is also every 100 ms, and the API already said "no" the first time it could be read, so "before" on Store P is before that API was readable. Store K and Store R are the same stores as in our Postscript test.
Swipe to see more
| Store / run | CMP | Marker | Marker time | Shopify sale_of_data "no", first read |
|---|---|---|---|---|
| O run 1 | OneTrust | OptanonConsent write, isGpcEnabled=1 | 4069 ms | 2371 ms |
| O run 2 | OneTrust | OptanonConsent write, isGpcEnabled=1 | 4336 ms | 2826 ms |
| O, no GPC | OneTrust | OptanonConsent write, isGpcEnabled=0 | 3621 ms | empty string at first read |
| K run 1 | Osano | OPT_OUT: ACCEPT | no later than 980 ms | 2807 ms |
| K run 2 | Osano | OPT_OUT: ACCEPT | no later than 1004 ms | 3245 ms |
| K, no GPC | Osano | OPT_OUT: DENY | no GPC marker | empty string at first read |
| P run 1 | No CMP | sale_of_data: "no" | no later than 1000 ms | same read |
| P run 2 | No CMP | sale_of_data: "no" | no later than 1176 ms | same read |
| P, no GPC | No CMP | sale_of_data: "" | no GPC marker | empty string at first read |
| R run 1 | Osano | OPT_OUT: ACCEPT | no later than 347 ms | 1732 ms |
| R run 2 | Osano | OPT_OUT: ACCEPT | no later than 264 ms | 1667 ms |
| R, no GPC | Osano | OPT_OUT: DENY | no GPC marker | empty string at first read |
The moment a banner processes GPC may depend on that store's configuration and script load order. We did not test either, so these times do not represent the CMP itself.
On Store O, about 70 ms after the cookie write, the page called Shopify setTrackingConsent at 4137 ms on run 1 and 4412 ms on run 2, with marketing false and sale_of_data false. Active groups after the GPC runs were C0001, C0002, and C0003. On the run without GPC they were C0001, C0002, C0003, SSPD_BG, C0004, and C0005.
In window.dataLayer, the Consent Mode default was ad_storage granted on Store O, Store P, and Store R, and denied on Store K. On the GPC runs the recorded gtag commands were default only. None was a consent update to denied. We only checked window.dataLayer. Other writes, including a Consent Mode path outside gtag and Shopify's own consent calls, were not in this check, so this record cannot show whether Consent Mode was updated some other way.
A separate page covers a GPC signal present but marketing still loads.
What was sent before the banner processed GPC?
Counts below are ad and marketing requests whose start time is before the marker. They include script downloads. The number in parentheses leaves out script, stylesheet, and document loads. Runs are not combined. We only checked whether requests were sent, not what they carried (for example an ad platform's restricted-processing parameters), so we cannot say whether any of them count as a sale or share.
Swipe to see more
| Store / run | Before (excluding script downloads) | Vendors and hosts |
|---|---|---|
| O run 1 | 62 (40) | Google Ads 23: www.google.com 9, googleads.g.doubleclick.net 5, www.googleadservices.com 4, cm.g.doubleclick.net 2, ad.doubleclick.net 1, the store's first-party sGTM subdomain (AW/DC gtag.js) 2. RTB House 9 (creativecdn.com). Amazon Ads 5 (c.amazon-adsystem.com and s.amazon-adsystem.com). Bing UET 4 (bat.bing.com). Kargo 4 (kds-pixel.kargo.com and storage.cloud.kargo.com). Ometria 4 (trk.ometria.com and cdn.ometria.com). Taboola 3. Postscript 2 (sdk.postscript.io). Reddit 2 (alb.reddit.com and www.redditstatic.com). MNTN 2 (dx.mountain.com and px.mountain.com). Xandr 2 (ib.adnxs.com). Teads 1 (p.teads.tv). The Trade Desk 1 (js.adsrvr.org). |
| O run 2 | 63 (41) | Google Ads 26: www.google.com 12, googleads.g.doubleclick.net 5, www.googleadservices.com 4, cm.g.doubleclick.net 2, ad.doubleclick.net 1, the store's first-party sGTM subdomain (AW/DC gtag.js) 2. RTB House 7. Amazon Ads 6. Bing UET 4. Kargo 4. Ometria 3. Taboola 3. Postscript 2. Reddit 2. MNTN 2. Xandr 2. Teads 1. The Trade Desk 1. |
| K run 1 | 2 (0) | www.googletagmanager.com (AW gtag.js) 1. static.klaviyo.com 1. |
| K run 2 | 2 (0) | www.googletagmanager.com (AW gtag.js) 1. static.klaviyo.com 1. |
| P run 1 | 5 (3) | Attentive 4 (cdn.attn.tv 3, a store-specific attn.tv subdomain 1). www.googletagmanager.com (AW gtag.js) 1. |
| P run 2 | 8 (4) | Attentive 7 (cdn.attn.tv 3, creatives.attn.tv 2, a store-specific attn.tv subdomain 2). www.googletagmanager.com (AW gtag.js) 1. |
| R run 1 | 0 | None |
| R run 2 | 0 | None |
On Store O the first ad request was at 1171 ms on run 1 and 1121 ms on run 2, about 3 seconds before OneTrust recorded GPC. Before Shopify read sale_of_data as "no", Store O had already sent 14 requests on run 1 and 23 on run 2.
How a CMP can hold tags before a click is a different check: CMP auto-blocking before Accept.
Did ad requests stop after GPC was processed?
The capture waited 20 seconds after DOMContentLoaded, with no click. The table counts ad and marketing requests from the GPC marker until that wait ended. How long that stretch lasts depends on when the marker landed, and this page does not turn it into one fixed duration after the marker. We recorded the counts and the vendors. We did not check which script sent each request. Again, we checked whether requests were sent, not their parameters, so we cannot say whether they count as a sale or share.
Swipe to see more
| Store / run | After (excluding script downloads) | Main vendors |
|---|---|---|
| O run 1 | 44 (16) | The Trade Desk 14, Google Ads 10, Amazon Ads 5, Bidswitch 4, Taboola 2, MNTN 2, Index Exchange 2, PubMatic 2, Bing UET 1, Xandr 1, Magnite 1 |
| O run 2 | 40 (12) | The Trade Desk 14, Google Ads 7, Amazon Ads 4, Bidswitch 4, Taboola 2, MNTN 2, Index Exchange 2, PubMatic 2, Bing UET 1, Xandr 1, Magnite 1 |
| K run 1 | 0 (0) | None |
| K run 2 | 0 (0) | None |
| P run 1 | 22 (16) | Google Ads 11, Attentive 8, Vibe 2, Mailchimp 1 |
| P run 2 | 19 (14) | Google Ads 11, Attentive 5, Vibe 2, Mailchimp 1 |
| R run 1 | 10 (5) | Google Ads 9, Klaviyo 1 (script) |
| R run 2 | 9 (5) | Google Ads 8, Klaviyo 1 (script) |
On Store O, after the page called setTrackingConsent with marketing false, the two runs still sent 44 and 40 ad requests.
How much did GPC change compared with a visit without it?
Each store has one run with GPC off, same homepage, same 20 second wait, no click. That run is only there to check that the marker moved with GPC. The table is the ad and marketing total for that visit.
Swipe to see more
| Store | Ad requests (excluding script downloads) | Vendors absent from both GPC runs |
|---|---|---|
| O | 153 (95) | AppLovin, Snap, Pinterest |
| K | 2 (0) | None. The same two script downloads. |
| P | 68 (55) | Meta, TikTok, Snap, AppLovin, Reddit, Spotify Ads |
| R | 22 (15) | Reddit, Amazon Ads, Pinterest |
On Store K, Osano's first read had MARKETING set to DENY on the GPC runs and on the run without GPC. The Consent Mode default on that store was ad_storage denied in all three runs. The run without GPC sent the same two script downloads, and nothing after.
What did Shopify's privacy API do on the store with no banner?
Store P shows no CMP. The first time Shopify.customerPrivacy.currentVisitorConsent() could be read, sale_of_data was already "no": no later than 1000 ms on run 1 and no later than 1176 ms on run 2. On the run without GPC that field was an empty string at the first read. Before the API was readable, the GPC runs sent 5 and 8 requests, Attentive and gtag.js. After it, Google Ads still sent 11 requests on each run, with Attentive, Vibe, and one Mailchimp request. We did not check which of these the store treats as a sale or share.
What the Customer Privacy API records, and what a customer-event pixel can still load, is on Shopify Customer Events versus the Privacy API.
What do California's official sources say about GPC?
California's AB 566, the California Opt Me Out Act, adds Civil Code section 1798.136. It says a business shall not develop or maintain a browser without a consumer-configurable setting that sends an opt-out preference signal. The section becomes operative on January 1, 2027. AB 566 bill text
Staff slides for the August board meeting listed GPC timing examples. Staff slides
Not legal advice.
On Store O, the first ad request went out at about 1.1 seconds. OneTrust recorded GPC at about 4.1 to 4.3 seconds.
What the CPPA's proposed rules (Nov 6 meeting) say
These are proposed, not in effect. The comment period is not yet open, and the notice of proposed rulemaking leaves its closing date as "[date]". They were item 9 on the agenda for the CPPA Board's November 5-6, 2026 meeting.
Not legal advice.
Section 7025 names GPC as an example of an opt-out preference signal. In the example at Section 7026(f)(3)(C) of the proposed text, Business WWW is an online retailer that uses a CMP and a tag management system (TMS). The example says: "Business WWW configures its website so that third-party advertising and analytics cookies and pixels that facilitate the sale or sharing of personal information are set before the CMP and TMS determine whether it has received the consumer’s GPC signal. Violet visits Business WWW’s website with GPC enabled in her browser and her request to opt out of sale and sharing is received as soon as Violet’s browser sends the initial request to access the website."
On Store O, OneTrust recorded GPC at about 4.1 to 4.3 seconds. Before that marker the two runs sent 62 and 63 ad requests.
How do Cookiebot and CookieHub say they handle GPC?
None of the four stores in this test ran Cookiebot or CookieHub, so this table comes from the vendors' own documentation, not from our runs. In Cookiebot, GPC changes behavior only for certain banner types. Cookiebot also says it does not honor the Do Not Track (DNT) signal. CookieHub 2.10.0, released October 29, 2026, lists the GPC changes in the notes under this table.
Swipe to see more
| CMP | Banner setup or visitor | What happens when GPC is on | Source |
|---|---|---|---|
| Cookiebot | Decline Only | Implied consent (clicking a link or scrolling) triggers an opt-out instead of an opt-in. | Cookiebot support |
| Cookiebot | Do Not Sell, banner not displayed | Full opt-out instead of opt-in. | Cookiebot support |
| Cookiebot | Do Not Sell, banner displayed | The "Do not sell" checkbox is pre-checked. The checkbox shows only when the banner is reopened. | Cookiebot support |
| Cookiebot | Multilevel or Inline Multilevel | Options that were pre-checked in the manager are switched to unchecked. | Cookiebot support |
| Cookiebot | Any other banner setup | No effect on consent or the banner. | Cookiebot support |
| CookieHub 2.10.0 | Installs not using the GTM community template | Live GPC overrides a stored choice. Consent Mode defaults are denied for DNT-tagged categories. Details below. | Source: vendor documentation, retrieved 2026-10-30, not tested by us |
CookieHub says version 2.10.0 (October 29, 2026) changes three things:
- A live GPC or Do Not Track signal now wins over a choice stored from an earlier visit, and the consent log marks a GPC opt-out as coming from GPC.
- With GPC on, Consent Mode defaults for implied-consent categories marked Used for tracking (DNT), such as
analytics_storage, are now denied instead of granted, and the category is not granted before the visitor answers. Do Not Track is treated the same way. - Consent events now record where they came from (a visitor action, a region default, a GPC signal, or closing the dialog), shown in the Dashboard consent log.
Installs using the GTM community template keep the template's defaults for a GPC visitor until they choose, because Google Tag Manager does not let a template read GPC; CookieHub says a later widget version will fix this.
Since 2.9, CookieHub says California visitors with GPC see a confirmation that the opt-out was received instead of the default dialog.
The $49 California audit records one visit with GPC off and one with GPC on.
A banner message saying GPC was received is the vendor's statement about consent. It does not show which ad requests stopped. To see that, compare a GPC-off visit with a GPC-on visit to the same URL, as in the tables above.
What we didn't test
One direct US exit, in Virginia. Two GPC runs and one run without GPC per store. Homepage only. No click, no scroll, and no product page.
Osano and Shopify marker times are upper bounds from a 100 ms poll. The OneTrust marker is the exact cookie write.
Before and after compare the request start time with the marker time. A request after the marker can continue a chain that started earlier, such as a cookie-sync redirect. We did not split those out.
We checked whether a request was sent. We did not read the fields inside it.
GPC in the browser was the header plus the JavaScript property. That should match Firefox or Brave with GPC on. We did not repeat the runs in those browsers. The browser steps are on how to test GPC in Chrome, Firefox, and Brave.
A few requests in the capture events had no Sec-GPC header: 1 of 764 and 1 of 761 on Store O, and 3 of 431 and 3 of 433 on Store R. The other requests on those runs had the header. Store K and Store P had the header on every recorded request.
These times come from four stores. They are not a ranking of CMPs, and they are not a statement that a store did something wrong.
How do you check GPC timing on your own store?
You can time the same gap in a desktop browser. Keep the header proof with the request list. The field list is on how to prove a storefront honored GPC. GPC timing is one of six items on the CMP checker checklist, next to Reject All, Accept All, and the stored choice.
- Turn GPC on before the load. Brave and Firefox have a setting. An extension can send it in other browsers. The free GPC checker reads the header your browser sent to ConsentProbe.
- Open a new private window. In DevTools, open Network and turn on Preserve log. Load the homepage and note the time of the first ad request.
- Open Application, then Cookies. For OneTrust, read isGpcEnabled=1 inside OptanonConsent. Or, in the Console, call
Shopify.customerPrivacy.currentVisitorConsent()and read sale_of_data.
Want this timing recorded on your store, run by run? See the cookie audit.
FAQ
Does GPC stop ad pixels on Shopify?
On these four stores, eight GPC runs, and one run without GPC each, the counts include script downloads. Store O sent 62 then 44 on run 1, and 63 then 40 on run 2. The run without GPC sent 153. Store K sent 2 then 0 on both GPC runs, and 2 without GPC. Store P sent 5 then 22, and 8 then 19. The run without GPC sent 68. Store R sent 0 then 10, and 0 then 9. The run without GPC sent 22. In total, the visit without GPC sent more of these requests than either GPC run on Store O, Store P, and Store R. Store K sent the same two requests with GPC on or off. Not legal advice.
How fast does OneTrust process GPC?
On Store O, OneTrust wrote OptanonConsent with isGpcEnabled=1 at 4069 ms and 4336 ms. That is one store. That store's configuration and script load order were not tested, so these times do not represent OneTrust itself.
Does Shopify's Customer Privacy API reflect GPC?
On Store P, Store O, Store K, and Store R, the GPC runs read sale_of_data as "no". On each store's run without GPC, sale_of_data was an empty string at the first read. This page does not describe Shopify's internal implementation.
Not legal advice
Not legal advice.
Sources
These links cover the platform and regulatory context used in this guide. Applicability still depends on the organization and jurisdiction.
- California Legislative Information, AB 566 bill text (Chapter 465, operative January 1, 2027)
- CalPrivacy, August board meeting staff slides
- Shopify Developers, Customer Privacy API
- Cookiebot Support, "Cookiebot and the GPC signal" (fetched 2026-10-29)
- CookieHub Docs, release notes 2.10 (retrieved 2026-10-30)
- CookieHub Docs, change log (retrieved 2026-10-30)
- CookieHub Docs, release notes 2.9 (retrieved 2026-10-30)
- CPPA Board meeting agenda, November 5-6, 2026 (agenda item 9)
- CPPA staff memo, agenda item 9 (November 6, 2026)
- CPPA notice of proposed rulemaking, opt-out preference signals
- CPPA proposed text, opt-out preference signals, dated 10.08.2026
Save a US-baseline technical record
After a DIY check, run a free US-baseline scan: one browser visit outside California, with cookies, requests, and screenshots stored as evidence. That visit does not run EU reject/accept or California GPC. EU, California, and Global 2 audits can be purchased from Billing after sign-in.