Findings · 2026-08-27 · By VSNARY | Emmanuel Orta · 2 views

Google's goto redirect is real. The referrer still can't show it.

Our one-session test saw no wrapped links. Checked against Google's confirmation, rank-tracker telemetry, other hands-on tests and the web specs, that sighting is an outlier. The argument about how the redirect works holds up.

Rewritten 2026-09-25. The first version of this page (27 August) reported one Denver browser session. We have since checked that test against Google's confirmation, rank-tracker telemetry, other hands-on tests and the web specs, 16 independent sources in all. Our sighting is one outlier among many. Our argument about how the redirect works holds up, and two parts of our own write-up were weaker than they read. Both are corrected below.

Headline takeaways #

Evidence key: Confirmed, Reported, Spec-backed, Inference (ours), Self-sourced (our own data, not independently checkable).

Finding 1: one Denver session saw no wrapped links #

On 26 August we searched from an ordinary Chrome session on a residential connection, read the result hrefs straight from the DOM and followed one click¹. We called it a sample, not a refutation, and it still is. The strongest rollout claims come from rank-tracking fleets on residential proxies, which is the traffic an anti-scraping measure would target first.

Finding 2: the sources disagree about who sees it, not whether it exists #

Google's only statement was generic: it has "a long history of deploying technical measures against evolving forms of abuse"². How much of the rollout people report depends on their vantage point:

Our reading (inference): the results fit a rollout gated by signals such as sign-in state and IP type. Symphonic's incognito-only result points to sign-in as one gate. We did not record whether our session was signed in, and that one detail could explain our null result.

Overstatement to avoid: some coverage says the rollout was "complete" by 26 Aug¹⁰. The vantage evidence above does not support "complete" for ordinary signed-in users.

Finding 3: the referrer argument, checked against the specs #

We said a landing page gets the bare origin https://www.google.com/ and never a path. Independent sources back each step:

• Google's results page sets its referrer policy to origin, so the browser sends only https://www.google.com/¹⁶.
• Under the origin policy, only the origin is sent, with a trailing slash¹².
• redirectCount returns 0 when a redirect is cross-origin¹⁴. The Fetch spec zeroes it if any hop in the chain is cross-origin¹⁵.
• Browsers default to strict-origin-when-cross-origin, which also drops the path on cross-origin requests¹³.

document.referrer = "https://www.google.com/"
referrer path     = "/"
redirectCount     = 0

What the destination page received in our session.

Digital Applied advised on 10 July: "As of today the expected result is zero hits — which is itself the data point."⁸ On-Page.ai likewise suggests watching reports for google.com/goto referral sources. Our point is that zero hits is not a data point, because the check can never fire.

The only record of the hop is on Google's side and in the browser's own network log, not in anything your page or server receives.

Where our first version was weaker than it read #

Its hop count could not count hops (spec-backed). We treated redirectCount: 0 as evidence that no hop happened. But google.com → destination is cross-origin, so the value is 0 either way. That is our own mechanism argument turned on our own test. The DOM href count is the evidence that actually holds.

Sign-in state was not recorded (inference). Symphonic reported the redirect in incognito only. Without knowing whether our session was signed in, the null result cannot be tied to IP type as opposed to account state.

"Invisible forever" assumes Google's policy stays put (inference). A redirect response can carry its own Referrer-Policy. The /goto hop could strip the origin as well; the first version covered this only as the "if goto ever strips referrers" case. On-Page flags the same risk⁵.

Self-sourced figures, commercial context. The "93% of arrivals sent no referrer" figure comes from our own five sites and cannot be checked independently. This page, like our others, links to our paid $749 fix service.

Finding 4: the one signal everyone agrees on #

Search Console counts the click at the result, before any hop. Analytics credits the session from the referrer on your site. If the redirect ever strips the referrer, the two counts drift apart. SUSO describes it as GSC holding steady while GA4 organic falls⁷. On-Page and Goodfellas Tech recommend the same reconciliation⁵¹¹.

Unconfirmed: SUSO said on 1 Sep that it "cannot confirm widespread cases of GA4 misclassification yet." Symphonic said no one had reported misattribution so far.

Rule these out first: consent banners, bot-filter changes, tagging gaps on some templates, and clicks that bounce before the page loads.

Finding 5: the cost lands on whoever reads results at scale #

The goto links cannot be decoded locally and Google refuses HEAD requests. Resolving one five-page ranking therefore takes 500 to 1,000 full requests, according to Nozzle⁴. Organic links are largely solved, but newer result features lag¹⁷.

DataForSEO, published 27 Aug 2026¹⁷, as reported by On-Page.ai⁵. Vendor-measured on its own traffic.

What to do this week: a baseline now beats a diagnosis later #

A measurement you start after the change has no "before" to compare against.

1. Record this week's ratio of Search Console clicks to GA4 organic sessions (same property, dates and country filter), and your organic/direct split.
2. Log the ratio weekly. If clicks hold steady while organic falls and direct rises, and rankings and impressions are flat, suspect the measurement layer.
3. Before blaming the redirect, rule out consent-banner changes, bot-filter changes and tagging gaps on specific landing paths.
4. Retire any alert or report that searches referrers for "google.com/goto". It cannot fire, so its clean result means nothing.
5. Ask rank-tracking vendors whether they resolve goto URLs, what their unresolved rate is for AI Overviews and local packs, and whether pricing or refresh rates will change.

What would change these conclusions #

• A published vantage panel. The same query run across many IPs, countries, sign-in states and devices, reading hrefs from the DOM, with each vantage stated. That would replace "uneven" with an actual rate.
• A referrer policy on the /goto response. If the 302 starts sending no-referrer, organic arrivals would show up as direct. The divergence check would then catch a real drop in attribution.
• Confirmed GA4 misattribution. Documented cases of GSC holding flat while GA4 organic falls, with other causes ruled out. None had been reported as of early September.
• Google documentation. Google has published no specification for goto paths. Official docs would settle scope and referrer handling directly.

Timeline #

Sources #

1. CrawlCheck: the original 27 Aug 2026 measurement (this page, first version)
2. Search Engine Land: Google confirms goto redirects, 26 Aug 2026
3. Search Engine Roundtable: rolling out goto, 26 Aug 2026
4. PPC Land: 500 to 1,000 requests per query, Aug 2026
5. On-Page.ai: google.com/goto is live, 28 Aug 2026
6. Symphonic Digital: What marketers need to know, 1 Sep 2026
7. SUSO Digital: rank tracking and reporting, 1 Sep 2026
8. Digital Applied: What happens to attribution, 10 Jul 2026
9. Around29: goto URLs and rankings, Aug 2026
10. WebEpex: redirects breaking rank trackers, Sep 2026
11. Goodfellas Tech: what it means for traffic, Sep 2026
12. W3C: Referrer Policy
13. MDN: Referrer-Policy header
14. redirectCount reference (MDN mirror)
15. w3c/navigation-timing issue #158, 2021
16. Jamie Wen: HTTP Referrer and Google's policy, 2020
17. DataForSEO: We already resolve 99.99% of URLs, 27 Aug 2026

Method: our original post read in full and cross-checked against 16 independent sources, searched 25 Sep 2026. Text labelled "our reading" or "inference" is ours, not a sourced claim. Superscript numbers link to the source list.

Questions this post answers

Is the Google goto redirect real?

Yes. Google confirmed the rollout on 26 August 2026. Coverage is uneven: rank-tracker fleets on residential IPs report near 100%, others saw it only in incognito, and our own residential Chrome session saw none of 34 links wrapped.

Can I detect the goto redirect from the referrer or my logs?

No. Google's results page sends only its origin, https://www.google.com/, as the referrer, and a cross-origin redirect chain reports redirectCount 0. Searching logs for a referrer containing /goto always returns zero, so a clean result means nothing.

How would a referrer-stripping redirect show up?

As a divergence: Search Console clicks hold steady while GA4 organic sessions fall and direct rises by about the same amount, with rankings and impressions flat. Record the ratio now so you have a baseline.

Has anyone confirmed GA4 misattribution from goto?

Not as of early September 2026. SUSO said it could not confirm widespread cases, and Symphonic said no one had reported misattribution.

Why does goto matter for rank trackers?

The links cannot be decoded locally and Google refuses HEAD requests, so resolving one five-page ranking takes 500 to 1,000 full requests per Nozzle. DataForSEO reported 52% of AI Overview SERPs and 25% of local packs still carrying an unresolved goto URL.

What was wrong with the first version of this page?

It treated redirectCount 0 as proof that no hop happened, but that value is 0 for any cross-origin chain. It also did not record whether the browser was signed in, which may explain why it saw no goto links.

Related findings

Comments

Comments are read before they appear. Nothing is published automatically, and no account is needed.

Writing about this? Facts, live figures and marks — every number on that page is dated and traceable to a scan.

All findings · The dataset · How the dataset works