Google's goto redirects: why search clicks now hop through Google first

Sometime in the past few weeks, most clicks on Google’s search results quietly stopped travelling straight to the websites they point at. They now hop through Google first, via a redirect URL on google.com/goto, before arriving at the destination. Google has confirmed the change is deliberate and rolling out broadly: a defence, in its words, built from “technical measures against evolving forms of abuse”.

The user barely notices. The measurement industry built on top of Google’s results is a different story. Here is what changed, who feels it, and what it says about where search measurement is heading.

What actually changed

Until now, a search result link pointed directly at the destination page. Under the new behaviour, the link routes through a redirect of the form google.com/goto?url=, where the destination is carried as an encoded value rather than a plain address, and Google’s server forwards the visitor on to the page. The result still displays the site’s URL beneath the title; what changed is the journey the click takes, not what the searcher sees.

The pattern was first spotted by search-industry observers in late June 2026 as a limited test. Roughly two months later, Google confirmed the deployment to search press, and rank-tracking vendors were reporting close to full rollout across the residential IP ranges they monitor.

Why Google is doing it

Google frames this as abuse protection, and the target is not hard to identify. An entire industry reads Google’s results by scraping them: SERP APIs that resell results data, rank trackers that check positions at scale, and, increasingly, AI systems that harvest search results to ground their own answers. Google has already taken one of the best-known SERP-scraping services to court this year, and the early rulings did not go all Google’s way. A redirect layer that obscures destination URLs behind encoded values is the technical answer to a problem the legal route has not solved.

There is a competitive layer to it too. Search results are among the most valuable structured datasets in existence for training and grounding AI systems. Wrapping the links raises the cost of taking that data without permission, at exactly the moment AI assistants have become Google’s most serious competition for the question-asking public.

Who feels it first

Ranked by exposure:

• SERP scrapers and rank-tracking tools. Most exposed, by design. Anything that reads Google’s results programmatically now has to detect and unwrap the goto pattern, and Google can change the encoding whenever it likes. Vendors are adapting, but third-party ranking data just became less certain and more expensive to produce.
• Analytics and attribution. The open question. A redirect inserted between the results page and your site could preserve the referrer, rewrite it, or strip it, and the answer determines whether organic traffic keeps classifying correctly. A stripped referrer would land organic sessions in GA4 as direct, one more blind spot in an attribution picture that was already degrading. To be clear: there is no evidence of that happening at scale, and Google has every interest in keeping its own attribution clean. But the mechanism now exists, and it belongs to Google.
• Searchers. Barely affected. The hop is close to invisible in normal use.

Notably absent from that list: Google Search Console, which reporting so far indicates is unaffected. That is worth sitting with. The one measurement surface that keeps working is the one Google provides directly.

The bigger pattern: measurement you do not own

Treated as an isolated technical change, goto is a minor story. Treated as a signal, it is a significant one. The platforms your visibility depends on are progressively closing the surfaces that third parties measure them through. Google has now put a wall between its results page and every tool that reads it from outside. The AI assistants went further from day one: there is no Search Console for ChatGPT, no referrer trail worth the name, no official way to ask which answers you appeared in.

The durable division is no longer organic versus paid, or search versus AI. It is measurement you own versus measurement you borrow. Search Console, your server and edge logs, your own analytics, and structured observation of what the answer engines actually say: those age well, because nobody can switch them off from the outside. Scraped SERP feeds and referrer-dependent attribution age badly, because they exist at the platform’s tolerance, and that tolerance is visibly narrowing.

This is the same conclusion we reached about measuring AI search visibility: when the platform publishes no rankings and grants no console, you measure by disciplined observation against a baseline, on infrastructure you control. Google’s results page is now drifting in the same direction.

What to do now

None of this calls for panic, and most of it calls for baselining before you need it.

• Record your organic-versus-direct split now. If a referrer change ever does bite, the symptom will be organic quietly reclassifying as direct. You can only spot that against a known baseline.
• Reconcile Search Console clicks against analytics organic sessions on a regular cadence. Divergence between the two is the early-warning signal, and Search Console is the side of the ledger this change does not touch.
• Check your server logs for google.com/goto appearing in referrers, which tells you how the hop behaves against your own traffic rather than in theory.
• Ask your rank-tracking vendor how they detect and handle the goto pattern. Their answer tells you how much to trust their trend lines through the transition.
• Shift structural weight towards first-party measurement. Server-side and edge-level logging, clean first-party analytics, and platform-provided consoles are the assets that survive this kind of change. Referrer-dependent, scraped and borrowed data are the liabilities.

How Morris McLane reads this

We treat platform-owned measurement as weather: useful, watched daily, and never load-bearing. Our AI search visibility work has assumed from the start that the engines publish no rankings and owe nobody a console, which is why AI visibility monitoring is built on structured observation: a fixed set of the questions that matter, run against the engines on a standing cadence, measured against a baseline we control. The goto rollout extends that logic to Google itself, and our measurement stack leans accordingly on Search Console, server-and-edge-level logging and first-party analytics rather than on scraped feeds.

The websites we build and manage apply the same principle at the infrastructure layer: managed AI-ready websites log what search engines, AI crawlers and assistants actually do on the site, at the edge, in first-party infrastructure, so visibility reporting does not depend on a referrer header surviving somebody else’s redirect.

If you want to know where your own measurement actually stands, the free AI visibility audit baselines how the AI engines currently see your organisation, using exactly the observation-based method this kind of platform change cannot break.

The short version

Google now routes search result clicks through google.com/goto redirects, confirmed as an anti-scraping measure and rolled out near-fully in August 2026. Users will not notice. Rank trackers and SERP scrapers will. Analytics attribution is unproven either way, so baseline your organic-versus-direct split and reconcile Search Console against analytics now, while the numbers are still known-good. The larger lesson is the durable one: platforms are steadily closing the surfaces third parties measure them through, and the measurement that survives is the measurement you own.