SerpApi said on September 5, 2026 that it had deployed a fix for the google.com/goto redirect links that Google began placing in front of search result destinations, and that its Google Search API is again returning the addresses of the sites themselves rather than a Google-owned wrapper.

In short

Google started routing search result links through an address on its own domain, so the link a data tool reads no longer contains the website it points to. SerpApi, which sells structured access to search results, says its engineering team worked out how to recover the real address and pushed the change across the affected result types. If you buy rank tracking or run a script that reads search listings, the field that carries the destination should behave as it did before, at least through this one provider.

What the company said it repaired

The post was published on the SerpApi blog under the byline of Hilman Ramadhan, the company's vice president of marketing and developer relations. According to SerpApi, Google recently widened the use of google.com/goto redirect addresses across search results, and affected listings pointed first at a Google-owned address that then forwarded the visitor to the final website.

The description of the fix is short. According to the company, the change briefly affected some of the URLs returned by its service, its engineering team identified the new behaviour, developed a fix, and deployed that fix across the affected Google Search result types. The company states that it is now returning direct destination URLs again.

That is the whole of the technical account. The post carries no date for when the breakage began, no date for when the fix shipped, no figure for the share of results affected, and no description of the method used to recover the addresses. It does not say whether recovery happens by following the redirect, by decoding the parameter, or by some combination that varies with the surface being parsed.

The field that stopped carrying an address

The more useful part of the post is its account of what did and did not break. According to SerpApi, the redirect did not affect visible result titles, snippets, rankings, or other parsed data. What it affected was one field.

The link field is the element of a structured search response that carries the address of the page a listing points to. Everything else about a result can survive a change to how the anchor is written, because titles and snippets are text in the page and position is a property of the ordering. The address is different, because it is the only part of a result that identifies the thing being ranked.

According to the company, businesses using search data to identify ranking pages or to monitor search performance could no longer reliably access or compare the actual addresses appearing in results. The post lists five categories of work it considers potentially affected: rank tracking and search analysis; citation and source extraction; AI agents and retrieval-augmented generation pipelines; link analysis and content discovery; and applications that store or compare destination addresses.

That list is worth reading slowly, because it spans two industries that rarely appear in the same sentence. The first, third and fourth items are the standard furniture of search measurement. The third is not. A retrieval pipeline that assembles citations for a generated answer needs the address of the source, and a wrapper that resolves only on click gives it a Google address instead. The same field underwrites a keyword report and a footnote in a chatbot answer.

Why one parser could not cover every case

The second half of the post addresses providers rather than end users, and it contains the only engineering detail the company offers.

According to SerpApi, the change required updates to how Google result links were decoded and normalised, and because the destination address could be encoded differently depending on the result type, a single parsing method could not resolve every variation.

That sentence describes a specific kind of work. A SERP is not one list. A single page can carry text results, sitelinks, image blocks, video carousels, top stories, product units and paid listings, each rendered by a different subsystem. If the encoding of a wrapped address differs between those subsystems, a parser has to be taught each variant separately, and the count of variants is not something an outside party can enumerate in advance. That is also why a fix announced as complete is a claim about the result types the provider has checked rather than a proof about all of them.

The post closes with the company's own framing of the episode. "Search engines are constantly evolving, and changes like this are one reason teams use SerpApi instead of maintaining their own Google Search parsers. We handle these changes behind the scenes to keep API responses stable, structured, and ready to use," the post states. That is vendor positioning rather than a technical claim, and it sits inside a post whose commercial purpose is not concealed: the page ends with an offer of a free tier at 250 searches a month.

What Google actually did, and when

The SerpApi account describes the redirect as a recent expansion. The public record is more specific, and it is longer than the word recent suggests.

Alex Greenland reported the first public sighting of rewritten result links on June 23, 2026. Brodie Clark, who publishes the SERP Alert newsletter, documented the pattern across both organic and paid listings on July 2, 2026, with a 302 response carrying the visitor to the destination. Barry Schwartz of Search Engine Roundtable could not reproduce the behaviour when he published on July 8, which is characteristic of a bucketed test rather than a shipped change.

Google confirmed the rollout on August 26, 2026. On the same day, Derek Perkins, founder and chief executive of the rank tracking firm Nozzle, published a chart covering roughly four months of testing and reported near total coverage across several residential IP providers. Google's statement named abuse rather than scraping, identified no category of target, and came with no timetable, no geographic breakdown and no technical documentation.

Two of Perkins's findings bear directly on SerpApi's claim. The encoded links cannot be decoded locally, which forces a provider to follow the redirect rather than read the address. HEAD requests, which would return the redirect target without the response body, are refused. A GET request that stops at the redirect works, but resolving every link on a single five-page ranking runs to between 500 and 1,000 requests, at which point rate limiting rather than bandwidth becomes the constraint.

Set that against the SerpApi post and a gap appears. If local decoding is not possible, a fix that restores direct addresses implies network calls that were not required in June, and those calls carry a cost and a rate limit. The post says nothing about either. Whether SerpApi found a decoding route that Perkins had not, absorbed the resolution cost, or narrowed the surface it resolves is not stated.

There is a second unresolved question in the same area. Dominic Woodman asked in the August thread whether the redirect is static per page, which would let a large operator build a lookup table once rather than pay per query. That question was not answered publicly, and it is the difference between a one-off engineering bill and a permanent tax on every tracked keyword.

The cost has already landed elsewhere

SerpApi is not the only measurement business to have published on the redirect in the past fortnight, and the other account is less sanguine.

SISTRIX, the German visibility platform, described the goto links as a change requiring an added-cost workaround, with founder Johannes Beus saying he expects further measures of the same kind. The company's newsletter framed the rollout as one item in roughly a year of adjustments that have made collecting search results at scale progressively harder.

The reference point for that year is the removal of a single parameter. Google eliminated num=100 on September 14, 2025, ending the ability to request a hundred organic results in one query, and Semrush confirmed a tenfold operational cost increase within a day. The sector response ran through late September, with AccuRanker limiting tracking to the top twenty results and SISTRIX discontinuing desktop results page updates. Ahrefs began restoring top 100 tracking on October 29, 2025 without disclosing the method, a reticence the current SerpApi post repeats.

SerpApi's own history with that episode is the closest available precedent for the announcement it has just made. The company built a workaround through its Google Light Fast API, and Google restricted that endpoint to three organic results within weeks. A fix that works on September 5 is a statement about September 5.

The structural point is that the two changes stack rather than replace one another. Removing num=100 multiplied the number of results pages a provider had to fetch. The goto wrapper multiplies the number of requests needed to interpret one page after it has been fetched. Both raise the cost of a tracked keyword, and they compound.

The litigation in the background

The company publishing this fix is the defendant in the case most directly concerned with automated access to Google search results.

Google sued SerpApi on December 19, 2025 under Section 1201 of the Digital Millennium Copyright Act, alleging circumvention of SearchGuard, the JavaScript challenge system it deployed in January 2025, and seeking statutory damages of 200 to 2,500 dollars per act. Judge Yvonne Gonzalez Rogers dismissed every claim on July 20, 2026, killing the general theory with prejudice and leaving only claims tied to specific licensed works. Google filed an amended complaint on August 10, 2026, dropping its Shopping and Maps theories and staking the surviving claims on three licensing contracts. SerpApi asked the court to end the case with prejudice on August 24, 2026, two days before Google confirmed the redirect rollout. A hearing is set for September 29, 2026.

A parallel action brought by Reddit against SerpApi, Oxylabs, AWMProxy and Perplexity AI began on October 22, 2025and also turns on SearchGuard. A separate post on the same SerpApi blog, published on August 31, 2026 under the byline of chief executive Julien Khaleghy, states that the company answered Reddit's suit and filed counterclaims of its own, asking a court to examine who Reddit shut out, what it promised, when it promised it and what it charged.

Schwartz connected the goto rollout to Google's legal position when he reported the confirmation, describing it as a likely attempt to make scraping harder for SerpApi. That is his assessment. Google's spokesperson made no reference to any case, and the docket supports the timing observation without establishing the motive. A court can decide whether reading a results page is circumvention. It cannot decide whether a link has to contain an address.

Why this matters for marketers

The first consequence is that a restored field is not the same as a restored cost base. Whatever SerpApi is doing to recover destination addresses, it is doing it against a page that no longer volunteers them, and the arithmetic Perkins published is the public estimate of what that costs. Providers that price resolution in will raise prices; providers that do not will absorb margin or reduce depth, which is what several did after September 2025.

The second is comparability across time. Ranking datasets assembled before June 2026, during the test window, and after the confirmed rollout may not measure the same thing, because what a provider captured depended on which infrastructure it used and whether it discarded affected pages or kept them. Nozzle discarded goto-only results pages during the test period. A dataset with holes in it and a dataset with wrapped addresses in it are different objects, and neither is the object that existed in May.

The third is that resolution is not a neutral act. Following a redirect touches Google's servers. Joe Manna asked in the August thread whether resolving a goto link registers as a click and whether click reports would begin carrying noise generated by rank trackers. The question was not answered. If resolution registers server side, the party doing the measuring becomes a source of the traffic being measured, and the wrapper doubles as a detection surface for automated clients.

The fourth reaches past tool vendors. A redirect hop between the results page and the destination adds a step controlled entirely by Google, on organic listings and on advertisement units through the aclk address. Referrer behaviour, latency and the ability of a person to see where a link goes before clicking it all pass through that hop. The same access question is being pressed on Google from the other direction, through Search Console reporting that now shows every site its appearances inside AI features while withholding the clicks and the queries.

Google has published no documentation for the goto parameter, has not said whether the rollout is permanent, and has not described what rate limits apply to resolution. One vendor says its output is back to normal. The conditions that made it abnormal have not changed.

Timeline

Summary

Who: SerpApi, the Austin, Texas search data provider, published the fix under the byline of Hilman Ramadhan, its vice president of marketing and developer relations. Affected parties include rank tracking vendors, competitive intelligence platforms, in-house measurement teams, retrieval pipelines that need source addresses, publishers and advertisers.

What: Google widened the use of google.com/goto redirect addresses in search results, replacing the destination address in the markup with a wrapper on its own domain. SerpApi states that the change briefly affected some URLs its service returned, that its engineers identified the behaviour and deployed a fix across the affected result types, and that direct destination URLs are being returned again. The post gives no dates for the outage, no share of results affected and no description of the method.

When: The SerpApi post was published on September 5, 2026. The redirect was first reported publicly on June 23, 2026, documented across organic and paid listings on July 2, and confirmed by Google on August 26, when Nozzle measured near total coverage across several residential IP providers.

Where: Google Search results pages, affecting the link field in structured search responses across multiple result types, with encoding varying by type.

Why: Google described the measure as protection against evolving forms of abuse without naming a target, and has published no documentation, no timetable and no rate limit guidance for the goto parameter. The change lands during active litigation between Google and SerpApi over automated access to search results, with a hearing scheduled for September 29, 2026.