Press review

Press review 08.11.2026: rank trackers show Google's redirect URLs, and PageSpeed scores move without site changes

Google is throttling how fast anyone can follow goto result links, so URL-level reports now show redirects instead of your pages. Add to that a block built around a single word on 7.6% of SERPs, and lab scores that drop with no change on your site.

Bartosz Bielawski

Bartosz Bielawski

· 6 min read
Share
Press review 08.11.2026: rank trackers show Google's redirect URLs, and PageSpeed scores move without site changes

Your rank tracker now shows a Google address instead of your page

DataForSEO and Serper both confirm that Google restricts how fast anyone can follow google.com/goto redirects from a single results page. The limit is applied per results page, not per IP address or query, so switching IPs or repeating the query does not help.

Once the limit is hit, Google returns a page with a "Continue" button, and the wait grows. With 18 links, the first five or six resolve right away, the seventh after about a second and the fifteenth after roughly 40 seconds, and DataForSEO recorded waits of up to 120 seconds. In High Priority and Live modes, only about half of the URLs resolve at all.

What to do: in every URL-level report, check whether the stored address is a google.com/goto link and compare it with the redirect target. A missing URL in a report does not necessarily mean a lost ranking.

Sources:

Google is mixing results: one in ten SERPs carries a block built around a single word

SE Ranking analysed 10,000 US desktop queries. In 84% of the flagged cases, the block of unrelated results was built around one word from the query, and 96.5% of those blocks started at position 9 or lower.

Nozzle tracks how the problem changes over time. In mid-October it affected 1.5% of results pages, and in the week starting 26 October it hit 7.6%. This is not a one-off experiment but something that comes back in waves.

What to do: before you blame an algorithm update for a drop, check whether a block of substitute listings is the cause. Compare the live SERP with impressions and clicks in Search Console for the same week.

Sources:

Your PageSpeed score can change even when nothing on your site has

PageSpeed Insights has moved to Lighthouse 13.6.0. That means your lab score can change without a single change to your code, because the simulation itself has changed.

Barry Pollard from Google explains that the new FCP and LCP simulation is more accurate, but it hits JavaScript-heavy sites harder. That is where the gap between the old and new scores will be widest.

What to do: record the Lighthouse version in every report and compare lab scores with field data. Without that, you will announce a gain that came from the tool, not from your work.

Sources:

Google, Microsoft and OpenAI now honour one shared robots.txt entry

The three companies announced on 30 October that they will respect a shared robots.txt control for AI training. It is based on an Internet-Draft by Kevin Kelley of Anthropic rather than on separate rules from each vendor.

The details matter. Google plans to treat the shared opt-out as an opt-out of Google-Extended when a site has no entry of its own, and that token also covers grounding in Gemini Apps. Microsoft will stop treating NoArchive and NoCache as training preferences after 1 March 2027.

What to do: choose deliberately between Google-Extended and the shared __usage_ai-training token, because their scope is not the same. If your site relies on Bing, put the migration date in your calendar.

Sources:

301 or 302 - this decides whether Google moves the signals to the new address

Google explained the difference plainly. With a 301, it passes ranking signals from the old page to the new one. With a 302, it assumes you will come back, keeps the old URL in the results and essentially ignores the new one.

Both types affect how similar pages are clustered, but only a 301 has a canonicalisation effect. For forms and payments, 307 and 308 matter because only they keep the POST data intact.

Google also calls cannibalisation a myth: it clusters similar pages and picks a canonical, and the real reasons pages go missing are lack of uniqueness and thin content.

What to do: during a migration, check that your redirects are server-side 301s, and on forms, check that you are not dropping the POST data. Meta refresh and JavaScript redirects are the worst options available.

Sources:

WordPress will not let you edit robots.txt, and AI crawlers do not read JavaScript

A list of eleven features that mainstream CMS platforms still lack starts with the basics. Out of the box, WordPress will not let you edit robots.txt, exclude pages from the sitemap or set noindex on a single page. A canonical tag requires a plugin or a developer, even though the tag has existed since 2009.

About 15% of sites still have no robots.txt file. AI crawlers, including those from OpenAI and Anthropic, do not execute JavaScript at all, so content rendered client-side is invisible to them.

The most dangerous bug on the list is a 404 template that returns a 200: the page says "not found", the server says "all good", and Google indexes the error page.

What to do: run through that checklist before your next build - robots.txt, sitemap, noindex, a 301 when a URL changes and a correct 404 status.

Sources:

Marketing and content

2,000 Google views and 12 calls - check who is actually finding you

An electrician in the US had 2,000 views of his Google profile in one month and 12 phone calls. The views are real, but most of them come from people who already know the company name.

Google counts each person once a day, so the number does not show demand. It shows recognition. The Searches tab in the profile's Performance section tells you more: it shows what people type to find the business.

What to do: open Performance, then Searches, and check whether the views come from the brand name or from services. If it is the brand, you have a visibility problem, not a conversion problem - and more reviews will not fix it.

Sources:

Mix one fitness video into ten business ones and the brand looks half fitness

Content about relationships gets around 10 times the reach of business content, and fitness gets 5-10 times. The effect can backfire: even one fitness video in ten makes the brand look split down the middle.

Reach matters more than the number of posts. The channel owner works out his mix from how he actually spends his time: fitness takes up 10-12 of the 168 hours in a week, so roughly that share should appear in the content.

What to do: work out the mix of your posts from the last 30 days and compare it with the share of time you really spend on each topic. If fitness is ten percent of your life but half your reach, cut it back before it reshapes the brand.

Sources:

Most ads only speak to people who already know the product

The four stages of awareness in Eugene Schwartz's inverted pyramid are problem aware, solution aware, product aware and most aware, which usually means existing customers. Most creators write content and ads for the last two groups only.

As a result, the ad never reaches people who have not yet put a name to the problem. Those people are drawn in by curiosity, not by a feature list or a comparison with competitors.

The author adds four ingredients of influence: status, power, credibility and likeness. Power matters most - whether your advice has worked before.

What to do: take your three best ads and check which stage of the pyramid they speak to. If all of them target product-aware buyers, write one for the problem stage.

Sources:

AI

Perplexity is selling agents that hand work to each other

Perplexity has launched a feature called teammate: separate task agents that pass work among themselves. A campaigns agent can hand a job to a legal agent to extend creator contracts. The agents can also message a colleague to move things along.

For now, it is a narrow feature. Only Max subscribers get it, at just under $200 a month, and only in the browser, with desktop and mobile versions promised later.

What to do: if you are testing agents, first write down which tasks should be handed over and under what conditions. Without that, you end up with four separate chats instead of one process.

Sources:

Read more