The American tax authority spent this week missing from Google for its own tax brackets and its own free filing pages. Nobody attacked it. Nobody penalised it. Somebody turned on bot protection a little too hard, and Google stopped being able to read the site.
That is the whole story, and it is coming for a lot of ordinary business websites in about two and a half weeks.
SEO and Search
The IRS fell out of Google because its own bot blocking shut Googlebot out
Over the past few days the IRS website and several other American government sites dropped heavily in Google. Lily Ray posted Sistrix data showing keywords the IRS ranked first for a month ago where it now barely ranks at all, including stimulus, free tax filing and tax brackets. Glenn Gabe found the same pattern across other government domains.
The tell was in the result itself. The IRS listing lost its favicon and its description, which is what you see when Google cannot fetch the page at all. Barry Schwartz reports the likely cause as bot blocking applied too aggressively, catching Googlebot along with everything it was meant to stop.
We covered the mechanism on 13 August, when Cloudflare announced new crawler defaults. What is new today is that a site the size of the IRS has demonstrated it live. Cloudflare classes Googlebot as a mixed purpose crawler, because the one bot both indexes for search and gathers data for AI training. From 15 September, any configuration set to block AI training blocks Googlebot too, including the older Block AI bots toggle that plenty of sites switched on last year without much thought.
Why this matters for your business: a great many South African sites turned on bot protection during the AI scraping panic, and almost nobody has looked at those settings since. Log in to your CDN or firewall today and find the bot rules. If AI training is set to block, add an explicit allow rule for verified Googlebot and Bingbot before 15 September. Then open Search Console, run a live URL inspection on your homepage and three money pages, and confirm Google can still fetch them. A 403 there is the difference between ranking and disappearing, and it produces no warning email at all.
Update: Google has now confirmed the goto redirect links we flagged yesterday
Yesterday we told you Google appeared to be routing result links through a google.com/goto redirect, with no comment from Google. Google has now confirmed it. A spokesperson said the company has “a long history of deploying technical measures against evolving forms of abuse”, which is corporate wording for an anti scraping move aimed at rank tracking tools and AI companies. This is now settled rather than suspected.
Why this matters for your business: expect rank tracking reports to get slower, patchier or more expensive over the coming months, because your tracker now has to follow every redirect to see where a result points. Ask your supplier what their plan is, and shift the weight of your reporting onto Search Console, which measures your own property directly and needs no scraping to work.
Social
LinkedIn users flagged a million posts as AI slop and those posts lost up to 40 percent of their views
LinkedIn gave people a button to report a post as “seems like AI slop”. More than a million people pressed it within about two weeks, according to LinkedIn chief product officer Hari Srinivasan. Views of content the platform then classified as AI spam fell by as much as 40 percent in some cases.
Be precise about the mechanism. One report does nothing, and LinkedIn says only very high volumes affect a post’s distribution. The notice you may now see in Post Analytics telling you readers think your content feels machine written is framed as feedback rather than a penalty. It is still a private note saying your writing did not land.
The more telling move is what LinkedIn did to itself. It has retired its own Enhance your post AI writing feature and replaced it with a proofreading tool built to preserve the writer’s voice.
Why this matters for your business: if your LinkedIn posts are drafted by a tool and pushed out unedited, your reach is already suffering and you will not see it in any dashboard. Readers are catching the pattern faster than the algorithm is. Write your posts in your own words, use AI to check spelling and structure rather than to generate the argument, and put one specific detail in every post that only somebody inside your business could know. That single detail is what separates a real post from a generated one.
Automotive
A read-across, not news: your dealership has more to lose from this than almost anyone
Nothing landed in dealer marketing overnight, so this is a read-across from the lead story rather than automotive news. It earns the space, because dealer websites carry a version of this risk that a plumber’s five page site does not.
A dealer group site is enormous, every vehicle in stock is a page, and those pages turn over constantly. That matters twice. Dealers get scraped harder than almost any other business, usually by competitors lifting stock and pricing, so website providers switch on aggressive bot protection as a matter of course. And because inventory pages appear and disappear all the time, a sudden fall in indexed pages looks completely normal on a graph. Your vehicle detail pages could be getting a 403 from your own firewall for a fortnight and it would read as ordinary stock movement.
Why this matters for your dealership: you probably do not control the firewall. Your website provider does, and their instinct is to block scrapers because that is what you asked for. Send them one question in writing this week: is verified Googlebot explicitly allowed through our bot rules, and will it still be allowed after 15 September. Then check it yourself. Inspect three live vehicle detail pages and one model landing page in Search Console and confirm each fetches successfully. Watch the indexed page count in the Pages report too, because a drop there with no matching drop in stock is your early warning.
What to do this week
- Open your CDN or firewall bot settings and add an explicit allow rule for verified Googlebot and Bingbot before 15 September.
- Run a live URL inspection in Search Console on your homepage and three pages that earn money, and confirm Google can fetch each one.
- Ask your rank tracking supplier how they are handling Google’s confirmed goto redirect links.
- Read your last five LinkedIn posts aloud and cut anything you would not actually say to a client.
- If you sell online, search ChatGPT for your own products, because its Shopping tab is now pinned to the sidebar and runs on a merchant feed you may not have set up.
- Dealers: get written confirmation from your website provider that Googlebot is allowed through, then verify it yourself.
If you would rather somebody else went through your firewall rules and Search Console before the middle of September, get in touch with Maple Leaf Agency.
Sources
- IRS Website Drops Heavily In Google Search, Search Engine Roundtable, 27 August 2026
- Report That Cloudflare AI Bot Blocking Prevents Googlebot From Indexing Sites, Search Engine Journal
- Block AI bots, Cloudflare developer documentation
- Confirmed: Google Search Rolling Out google.com/goto Tracking Parameters, Search Engine Roundtable
- LinkedIn’s ‘Seems like AI slop’ button has already been used 1 million times, PCWorld
- Shopping Pinned To ChatGPT Sidebar, Search Engine Roundtable