Google now sends search result clicks through a redirect it controls

TL;DR: Until recently, clicking a Google result took you straight to the website. Google has changed that. The link now points at an address on Google’s own domain — google.com/goto?url= followed by a scrambled version of the real address — and that page bounces you onward to the site you wanted. You still land in the same place, a fraction of a second later. Google confirmed the rollout on 26 August and called it a measure against abuse. The people it makes life hard for are the companies that read Google’s results automatically and sell that data — which is where almost every rank-tracking report you have ever sent a client comes from.

First, what a redirect actually is

A redirect is a stop-over. Instead of a link taking your browser directly to example.com/shoes, it takes you to some other address first, and that address immediately replies “not here, go to example.com/shoes”. Your browser follows. You usually notice nothing except a flicker in the address bar.

Redirects are ordinary plumbing. Every shortened link works this way. Every “click here” in a marketing email works this way, which is how the sender knows you clicked.

What is new is where Google has put one: on the organic results themselves, the ten links that have historically pointed straight at the destination.

What Google actually said

Barry Schwartz at Search Engine Land asked Google about it and got a statement back: “We have a long history of deploying technical measures against evolving forms of abuse, and we regularly take steps to protect our services and users.”

That is a confirmation, not an explanation. Google did not spell out which abuse, and it did not publish anything in its documentation.

On scale, the useful number comes from Derek Perkins, who runs the rank-tracking company Nozzle. He reported seeing the change at “nearly a 100% rollout across several residential ip providers” — meaning it showed up on ordinary consumer internet connections in several places, not just in a small test. Search Engine Roundtable had spotted the format in testing earlier in the summer, so this has been coming for a couple of months.

Who this is aimed at

Rank tracking is not magic. When a tool tells you a client sits at position 6 for “emergency plumber leeds”, a machine typed that query into Google, from a specific location, and read the page that came back. That reading step is called scraping. Some tools do it themselves; many buy the data wholesale from a supplier.

A scraper reads the HTML of the results page and pulls out the links. When those links were plain URLs, the destination was sitting there in the markup, free to take. Now the markup holds an encoded string pointing at google.com/goto, and getting the real address out of it means actually following the redirect — one extra request per result, per keyword, per location, per day. At the volume these companies operate, that is a serious cost increase, and Google is the one metering it.

This sits alongside the legal route Google is already taking. It sued SerpApi, one of the larger wholesalers of search-results data, and refiled a narrower version of that case in August. We wrote about what that lawsuit means for the tools your reports run on. Now there’s a technical squeeze arriving at the same time as the legal one.

What breaks, and what doesn’t

Be careful here, because there is more speculation flying around than measurement.

What Search Engine Land reported: regular searchers are barely affected, apart from losing the preview of the real URL in the status bar when hovering a link. Google Search Console data should be unaffected — that is Google’s own record of impressions and clicks, taken from inside Google, so a redirect on the way out doesn’t touch it.

Third-party rank trackers are the ones with a problem, and how much of one depends on how their supplier is built. Some will absorb it quietly. Others will get slower, more expensive, or patchier, and you may see that as gaps in a report rather than as an announcement from the vendor.

Analytics attribution is the open question. Because the hop happens on google.com, the referring domain your analytics sees should still be Google — but nobody has published measurements on how this behaves across browsers and privacy settings yet, so treat confident claims either way as guesses for now. If your organic numbers move oddly over the next few weeks, this is worth ruling in or out before you go looking for an algorithm update.

What to do this week

Check your rank-tracking data for holes. Not the trend line, the raw rows: missing dates, keywords that silently stopped reporting, positions that froze. A degraded scraper often looks like stability rather than an error.

Then ask your vendor directly whether they source their own results or buy them, and what their plan is. It’s a fair question now, and the answer tells you how exposed your reporting is.

And weight Search Console more heavily in client reporting than you did last month. It has always been the more trustworthy of the two — it is Google’s own count rather than a reconstruction from the outside — and it is the one measurement channel this change cannot touch.

The direction here is simple enough. Data about your visibility, gathered from outside Google, is getting harder and more expensive to obtain. Data about your own site, and about what AI answer engines actually say when asked, is not.

That second half is what Preferium is built on. It measures four AI engines on every plan — ChatGPT, Claude, Perplexity and Gemini — plus Google AI Overviews and AI Mode tracked separately, which is six answer surfaces checked directly rather than inferred from a scraped results page. On the site itself, 47 automated checks crawl every page and score it out of 1000, and the system fixes what it finds and re-checks the live page in a real browser afterwards. None of that depends on reading Google’s SERP HTML. More on how the system works.

Key takeaways

• Google confirmed on 26 August that organic result links are being rewritten to google.com/goto?url=[encoded], which then redirects to the real destination.
• Google’s only stated reason is protecting against “evolving forms of abuse”. No documentation, no detail on which abuse.
• Derek Perkins of Nozzle reports nearly 100% rollout across several residential IP providers, after a testing period earlier in the summer.
• The cost falls on scrapers: the real URL is no longer sitting in the results page markup, so each one has to be followed.
• Search Console is unaffected. Third-party rank trackers may degrade quietly rather than fail loudly — audit your raw data for gaps, not just your charts.
• Analytics referrer behaviour is unmeasured so far. Rule this change in or out before blaming an algorithm update for odd organic numbers.