Google confirmed it is rewriting search result links through google.com/goto, an anti-scraping measure that breaks domain matching for every tool that reads its results, and Google's Search Central blog broke its August silence with a site reputation policy update that names the four criteria it uses to judge third-party content.

A result link that used to point at the destination page now points at google.com/goto?url= followed by a hash, a server side redirect that then forwards the visitor to the real page. The encoding is custom and base64-like, and neither report was able to decode it. Google confirmed the rollout with a spokesperson statement, quoted in full as "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." Derek Perkins of Nozzle described it as "nearly a 100% rollout across several residential ip providers" on 26 August, after roughly two months of testing. The earliest practitioner sighting on record is 23 June 2026.

DemandSphere, a rank tracking vendor and therefore self interested, stated the consequence most clearly on 27 August: without a countermeasure every destination URL appears to resolve to google.com, domain matching fails, and historical series break at the rollout date. It says the change applies to organic result links and to SERP features. Neither report states whether AI Overviews or AI Mode citation links are wrapped the same way, and that question is unanswered in every source found today, so treat it as unknown rather than as no.

The post itself is about enforcement geography. Following discussion with the European Commission, Google is adjusting how it applies the policy inside the European Economic Area. From 30 August a manual action affects search results for users outside the EEA as before, while inside the EEA the impact does not apply and the affected section may instead be separated in Google's systems so it ranks independently from the rest of the site over time. Site owners are still notified in Search Console, can still file a reconsideration request, and eligible sites gain a route to mediation. Google keeps its objection on the record, saying it remains concerned "that an overbroad application of the DMA could prevent us from addressing real threats to the integrity of our search results."

The part that matters outside Europe is half a sentence, that Google is also "clarifying the criteria we consider when applying the policy." There are four, and Google says explicitly that "not one of these factors is either necessary or sufficient on its own to determine" a violation: how the content is presented, meaning whether design, formatting and typography match the host; the quality of the content relative to the host's own standard; its stated or implied authorship; and whether the content appears on multiple other sites. The documentation also lists what is not a violation, including wire services and press release distribution, syndicated news, forums and comment sections, columns and opinion pieces, and third-party content genuinely meant for readers rather than for rankings.

Method, stated on the page: a representative panel of 1,002 US adult consumers fielded via SurveyMonkey, with the AI specific questions answered by the 455 respondents who had used an AI tool for a local business recommendation in the previous twelve months. Self reported behaviour, so it measures what people say they do.

The movement is 6 percent in 2025 to 45 percent in 2026 for using AI to find a local business, which makes AI the third most used discovery channel behind Google and Facebook, ahead of Yelp and TripAdvisor. Google's own share of local business discovery fell from 83 percent to 71 percent across the same year. By tool, ChatGPT is at 31 percent of consumers and Google AI Mode is second at 23 percent, with Gemini, Copilot and Claude below them. On trust: 63 percent of AI users say they trust the recommendations, 10 percent distrust them, and 88 percent fact-check what they are told, with 42 percent saying they always verify against traditional review sites.

Method as stated: roughly 500 prompts across 3 countries and 6 cities, covering lawyers, accountants and dentists. The article does not publish the prompt list, the engines tested or the date range, so it is weaker evidence than its outlet suggests and should be read as a practitioner study rather than a dataset.

Four findings worth recording. Even small amounts of earned press coverage moved recommendations, and the effect was strongest in smaller and less competitive markets. Business names carrying a geographic signal were connected to location queries more readily. Sector matters: lawyers tracked closely to conventional search rankings while accountants diverged most widely. And external coverage functions as a filter, with the assistants appearing to pre-select on third-party confidence signals before composing an answer.