Google's /goto redirect: another reason to trust first-party data
Google now hides destination links behind an encoded /goto redirect, making SERP scraping slower and more expensive. Here is what changed and why your own Search Console data is unaffected.
If your rank tracker has been reporting broken links, missing results or odd gaps since late summer 2026, there is a concrete cause. Google has rolled out a new way of writing the links in its search results, and it is specifically designed to make automated scraping harder.
What actually changed
Google confirmed in September 2026 that it had deployed /goto URL redirects across search result links, after several months of testing (searchengineland.com, September 2 2026). Instead of the result linking straight to the destination site, the link now points to a google.com/goto endpoint with the real URL hidden inside an encoded parameter.
That encoding uses a Google-specific format that cannot be decoded on your own machine. To find out where a result actually points, a tool has to send a full request to the redirect and read where Google sends it back. Google does not allow the lightweight HEAD requests scrapers would normally use for this, so every single link now requires a full page request to resolve.
Why that matters for rank trackers
The practical effect is a tax on scraped data. Industry write-ups estimate that resolving the links for a single five-page ranking check can now require on the order of 500 to 1,000 extra HTTP requests (webepex.com, September 2026). That means higher proxy costs, more latency, and more ways for the data to come back incomplete.
Vendors have reacted differently. Semrush and SE Ranking said they were largely unaffected or adapted quickly, while SerpApi and Ahrefs acknowledged disruption and worked on fixes (searchengineland.com, September 2026). If a tool has not adapted, its users can see skewed positions, missing rows or broken destination links.
There is a second, quieter effect. Because the redirect runs through Google, the referrer information that analytics relies on can be stripped, which can push some organic sessions to show up as direct traffic in Google Analytics. Search Console reporting, which works at the search level rather than from the click, is not affected by this.
This is not an isolated event. It follows the removal of the num=100 parameter in 2025 and the poisoned-results behavior of 2026. The direction is consistent: scraping the SERP keeps getting harder, slower and less reliable.
Why first-party data sidesteps all of it
The /goto redirect only affects tools that read Google’s results pages from the outside. Search Console does not read the results page at all. It is first-party data from Google’s own logs: the real impressions, clicks and average positions Google recorded for your site. There is no link to decode and no page to scrape, so an encoded redirect changes nothing about it.
The advice analysts are giving right now is the same advice this change keeps reinforcing: when a third-party tool disagrees with Search Console, trust the first-party record and look at the trend over time rather than a single day.
The honest caveat
First-party data is not a complete substitute for scraped tracking. Search Console positions are averages weighted by impressions across devices and locations, not a single point-in-time rank, and it has no competitor data. Scraped trackers still have legitimate uses for competitor and local visibility. The point is that your own Google data is the most authoritative record of how you actually performed, and every one of these Google changes makes that record more valuable relative to the scraped alternative.
What SearchVault does with it
SearchVault is built entirely on first-party Search Console data, so none of the /goto disruption touches it. It keeps every month of your history permanently, tracks daily rankings and performance from Google’s own numbers, and turns them into plain next steps, so when the scraped world gets noisy you still have a clean, trustworthy baseline to work from.