First in Generative Engine Optimization (GEO)

Google cares about the helpfulness of content, not whether it’s AI-generated. The problem is that AI-generated content is not unique or helpful because it’s built from data sets of content others have already published. (This is also why companies with copycat blogs written by humans are suffering from HCUs.)

Content, engagement and conversion experts warned that using AI to generate content was fraught with peril and that Google didn’t need to explicitly search for and penalize AI content for website owners to be hurt by AI-generated content.  The unoriginal content and plummeting engagement signals would lead to that outcome on their own. But, human nature being what it is, companies hit the “easy button” in droves and not only fell into the AI-generated-pablum trap but also compounded the problem by making even more essentially the same content (if HCUs have hit your company, see our FAQ note below).

As for confusion, we believe that confusion over AI’s impact on HCUs stems from two primary factors. First, generic human-generated content (e.g., copycat blogs) have been de-ranked along with de-rankings associated with AI content (unhelpful content is unhelpful content,  period). Second many SEO firms haven’t adjusted to a world in which content genuine rules the day. Their orientation and understanding of what is or isn’t valuable is inverted to this new reality and they have no workarounds to succeed when they don’t have relevant or meaningful content.

We encounter this query regularly, sometimes from companies that have experienced breathtakingly steep drops in website traffic following HCUs. Sometimes, the traffic drops ladder downward with successive HCU updates.

Other times, a single update creates a plummeting traffic cliff. These aren’t directly related to generative search, per se, but they reflect the reality that Google has to compete with generative engines.

Also, if you rank poorly in traditional search, your chances of ranking well with GEO are low.

These are content problems. You need to develop original, relevant and helpful content. There are methods of engineering your content for this new reality that don’t necessarily mean finding a Hemingway-in-a-haystack SME or deploying massive content teams.

In fact, if you do it right, you’ll do better with a lean, tech-enabled team. (Work with a benchmarking firm at our 10-year mark revealed that our boutique revenue marketing agency has influenced more than $40 billion in revenue across generation, GTM and retention activities.) But if you’ve been walloped by HCUs, you likely need to reorient yourself from stuffing “easy-button” filler into search strategies and automations, and toward quality content with tech-enabled expansion and distribution.

Note: As Google adjusts its business model to (external and internal) competition from AI competitors, its HCU updates sometimes create wild swings. There are cases wherein companies have been de-ranked and recovered. If you’ve been suddenly de-ranked (the overnight plummet, not the successive decline across multiple HCUs) and want to gauge the likelihood of it being a blip with recovery, look at your content.

Is it original, helpful and relevant, or is it generic, copycat or AI-generated from the web (instead of original documents)?

While AI can help with content ideation, a successful GEO strategy requires human expertise, oversight, and original insights. AI can be powerful in distribution, not origination. Getting your horse-and-cart order right helps with engagements and conversions, too. 

GEO success combines traditional metrics (rankings, traffic) with new measurements like snippet inclusion, proof-point visibility, and conversational alignment testing. You still need to track conversions and keep an eye on SEO performance. SEO and GEO aren’t either-or propositions. You need them both. A good generative engine optimization (GEO) agency will be able to provide you with AI search analytics that help you identify opportunities, see your competitive placement, and track your progress.

At this point, many of the tools you currently use, like Google search console and CRM integrations, can help you obtain many of the metrics you need. Additionally, some of the SEO software suites (SEMRush, Ahrefs, SimilarWeb, etc.) are starting to offer rudimentary visibility data.

You might expect us to endorse AI detection software since we have built a successful revenue practice around high-value content. We don’t. There’s a lot of misinformation about this subject, so here’s a rundown on why.

When AI detection software first emerged, we found it buggy but potentially promising with iteration because it flagged poorly written content as likely being AI-generated.

In testing, our content team took high-value content and rewrote it poorly to make AI detection software think it was AI-generated— an incorrect result but a screen against lousy content (hence why we thought it might have promise if it could get actual detection down).

But as of the time we updated this section of these FAQs, it also thinks too-well-written content is AI-generated—to the point that using grammar-checking software (an unquestionable best practice that helps with humans and Google) will generate erroneous flags.

You don’t have to take our word for it. OpenAI is on record saying what most content specialists know all too well – the detectors simply don’t work:

“To elaborate on our research into the shortcomings of detectors, one of our key findings was that these tools sometimes suggest that human-written content was generated by AI.”

In our own firm’s most recent test, we swayed the market-leading tool’s (which boasts self-proclaimed accuracy of 99% despite a wall of disclaimers) assessment of an entire 1,400-word document’s AI probability by more than 50 points by simply rewriting a catchy CTA. The snappy writing apparently colored the software’s analysis of the rest of the content on the page because replacing it cleared other sections that it previously found suspect. Needless to say, we kept the catchy CTA to capture conversions instead of making the detector happy.

Perhaps more importantly, the premise of AI detection software – that it protects you from Google detecting and de-ranking AI-generated content – is incorrect.

Google has very deliberately explained that it cares about helpfulness, not origination. Here again, it’s easy to become snared in the confusion over AI’s impact on search results (as discussed in FAQs above). AI-generated content doesn’t generally perform well because it has nothing new and insightful to say, not because Google thinks it’s AI-generated.

Our advice? Use grammar and plagiarism checkers, have your SMEs check your content for helpfulness and original insight – which is what Google’s actually looking for – and move on to whatever’s next on your list. Chasing detector clearance is a waste of time at best, and if you’re good at what you do, it might even degrade your work (and results).