Why Unedited AI Content Triggers Google Penalties

In this week’s episode of Voices of Search, we spoke with Kaspar Szymanski, founder and senior director at SearchBrothers and a former member of Google’s search quality team, where he investigated web spam and applied manual penalties. Google recently completed its third spam update of 2026, and Kaspar sees more going on than the AI headlines suggest.

Our conversation covered why the August spam update and a wave of manual actions landed together, what separates smart AI use from penalty-inducing automation, and how to approach prevention and recovery.

Key Takeaways From This Episode:

• The August spam update and a wave of manual actions landed together, and some sites were hit by both
• Google is indifferent to how content gets produced, and sites get flagged for low-quality output, often because human oversight was missing or inadequate
• Unedited AI content is easy to spot, and everyday users are getting better at spotting it too
• Prevention is far cheaper than remediation, and a periodic audit covers technical, policy, and backlink signals
• After a drop, calm investigation comes first, and a rushed reconsideration request makes recovery harder
• Diversified traffic and a recognizable brand limit the damage when a penalty hits anyway

Two Enforcement Waves Landed at Once

The August spam update rolled out globally between August 18 and 21, and Google didn’t name a specific target. Early coverage focused on AI, and industry reports suggest sites with mass-generated AI content may have been among those hit hardest.

Kaspar’s team sees a second effect that gets less attention. Over the last six weeks, Google has also pushed a wave of manual spam actions aimed at violations that went unnoticed for a long time. Some sites are affected by both. The two work differently:

• Spam update: Algorithmic and broad, with no notification. Google puts heavier emphasis on methods it doesn’t approve, and borderline sites tend to lose out
• Manual action: Targeted at specific violations such as manipulative link building, cloaking, or low-quality content, applied to a section of a site where possible, with a notice in Search Console. It takes a long time to expire, and waiting it out isn’t an option, which he compared to an aching tooth

He doesn’t think the overlap is a coincidence. Spam detection is both an art and a science, he said, and if algorithms could handle it alone, there wouldn’t be a web spam team. He also called the summer timing a blessing in disguise, since remediation takes time and a hit in mid-November would land in the middle of Q4 sales.

“This is all fixable,” he said. “It may take a little bit of legwork, but it can be done.”

Quality Control Is the Missing Piece

Most of the conversation around the update centers on AI. Kaspar, drawing on his years inside Google, said the company is “indifferent with regard to how content is generated.”

Automated content has existed for as long as Google has, and large volumes of templated landing pages are fine as long as they add value. He pointed to sites that mashed up content from several sources a decade ago and created something readers couldn’t get elsewhere. That raised no quality concerns.

What AI changed is who can publish at volume without an editorial process behind it. The problem behind most affected sites, he said, is the absence or inadequacy of quality control and human oversight.

That shows. Unedited AI content is easy to spot, and everyday users are getting better at spotting it too. A growing group of people doesn’t like it, and some YouTube creators now state that no AI was used in their videos.

“There is nothing inherently wrong with using AI,” he said. “It can be used if it’s being used wisely.”

The risk climbs significantly when AI generation takes over the editorial process completely. In his view, the recent moves don’t target AI specifically. Google is targeting what it considers low-quality content.

Oversight Looks Different for Every Business

Tyson pointed out that production has become easy, which leaves vetting as the harder part. Kaspar agreed that validation is the weak spot of most AI tools. Anyone who has run a query and received two opposite answers has seen the problem.

The goal, he said, is content created for humans, not content produced at scale to fill quotas. What that looks like depends on the business:

• News publishers: Time is of the essence, so crawl budget and server response matter, but Kaspar has seen publishers replace staff with AI uncritically. Critical thinking is something AI can’t supply
• Retailers: Properly marked-up product pages and genuine user reviews, with fraud kept out, can matter more than a large volume of new text
• Event sites: Added value can be data, such as which stadium sections are popular, which helps a parent pick a quiet area to watch the game with kids. AI can generate that kind of information as long as it helps the user, not Googlebot

Tyson noted that small sites can review everything by hand, while larger operations have to work out how to scale that scrutiny.

“There is, I’m afraid, no silver bullet short of saying human oversight is critically important,” Kaspar said.

Prevention Works Like a Garage Check

That oversight extends past individual pieces of content. Kaspar’s view is that SEO deserves a line in the budget and in policy compliance, but it shouldn’t fill a CEO’s day. A better product or service matters more, since SEO helps a good business sell more and can’t fix one that’s missing something customers want.

He compares the alternative to a periodic vehicle check. A website is a living organism, with shifting Google policies, aging technology, and frequent edits, and mergers or domain and CMS migrations can change signals considerably. An annual audit should cover:

• Technical signals
• Policy compliance
• Backlink signals, which he called a diminishing factor that can still be a risk

It should also confirm the business model isn’t fundamentally at odds with Google’s guidelines. If it is, that’s the owner’s right as long as it’s legal, he said, but there should be a contingency plan.

Cost is the usual objection. Kaspar’s response is to ask how much the lost visibility is costing every month. In his experience, one to three months of waiting with a lingering penalty costs more than an audit and a remediation effort, and prevention costs less than either.

“Prevention is much, much cheaper,” he said. “It can be planned in advance at a time that is convenient.” He stated.

When the Drop Comes, Investigate Before You Act

Prevention can still fail. Kaspar’s first instruction for anyone hit is to stay calm, since agitation won’t help, even in conversation with a Google representative. From there, he follows an order of operations:

• Check Search Console for a manual action. If there isn’t one, the drop likely ties to the spam update
• Save and preserve server logs. Many sites purge them after 30 days, before there’s enough data to see patterns
• Crawl the site to see what it actually contains, including crawler traps and lean landing pages nobody knew existed
• Verify backlinks beyond the samples that third-party tools and Search Console provide
• Bring in a trusted third party, since in-house teams are often too attached to the site to find its flaws

Technical, content, and off-page signals affect each other, so a single finding doesn’t settle the case.

“Finding one that may be an issue isn’t really quite a smoking gun,” he said. “You need to dig deeper to understand how the system is being perceived by Google.”

Server logs deserve one caution. Increased crawler activity before a drop correlates with the problem areas in most cases, but it can also mean the server got faster or demand rose. It’s useful in hindsight, not as a warning light, and he asked listeners not to check logs daily.

Recovery from an algorithmic hit comes down to crawl budget. Fast servers and strong internal linking encourage deeper crawling, and for most sites that means one to three months. Nobody has to wait on a reply from Google, and cleaned-up sites often come back stronger.

Rushing a Reconsideration Request Backfires

For manual actions, Kaspar said the biggest mistake is filing a reconsideration request right away, on the assumption that it holds a place in a queue. Rejected requests stay on record.

“Every consecutive reconsideration request that has been rejected makes the task at hand harder because there is a loss in trust,” he said.

Deleting parts of the site doesn’t remediate anything either. Analyzing the cause first typically shortens the time a penalty stays in place.

Spread the Risk Before the Next Update

How hard a penalty hits depends on how much of the business rides on one source. Kaspar’s advice is not to keep all your eggs in one basket. Relying on Google or Bing alone is dangerous, and traffic from other sources can keep revenue flowing while a site works its way back into Google’s good graces.

Brand building is the stronger version of that. It takes time and money, he said, but it makes a business less dependent on organic traffic, and direct type-in traffic is something nobody can take away. He keeps the basics simple:

• A real logo from a real designer, not one bought for a few dollars
• Knowing your audience and engaging with them where possible, well beyond Reddit
• Recognition that lifts click-through rates, which helps SEO down the line

Some operators run several brands as a risk strategy. Kaspar doesn’t think that’s a bad one, but it only works if the brands truly differ, with separate accounts, servers, CMSs, and logos.

“That’s not diversification. That’s a network,” he said, “and networks, once identified, are easily tackled.”

Core updates call for the same mindset. Google releases them more often than it used to, so sites can be hit and can recover more often, and nobody can force a release. Kaspar calls it SEO weather. Ignore what you can’t control and focus on the signals you can, and stay honest about recovery timelines, since overselling them erodes trust.

Human Oversight Is the Best Penalty Protection

Everything in this conversation traces back to one point. Google cares about what ends up in its index, not how it got there. AI gave people a way to skip the editorial process, and skipping it is what draws penalties. The same discipline applies to the whole site, through periodic audits, preserved server logs, and traffic that doesn’t depend on a single source.

Kaspar’s main message is that every one of these situations is fixable, and that prevention costs far less than recovery. 

Teams that keep a human in the editorial process and audit their sites on a schedule will rarely need the recovery playbook.

Voices of Search is a daily SEO and content marketing podcast hosted by Jordan Koene and Tyson Stockton. The show delivers actionable strategies and data-driven insights to help marketers navigate the ever-evolving world of search engine optimization and content marketing. New episodes air weekly, covering everything from technical SEO to AI discovery, featuring industry leaders and practitioners sharing real-world frameworks and proven tactics.

Subscribe to Voices of Search on Apple Podcasts, Spotify, or your favorite podcast platform. Follow Previsible on LinkedIn for updates and subscribe to the VOS YouTube channel for video episodes and clips.