SEO Expert Exposes Google's AI Search Guidelines (The Truth)
An 18-year SEO veteran says Google's AI Search Guidelines are quietly steering you away from the exact tactics Google uses on its own sites.What Google's AI Search Guidelines Actually Say
What if the most-shared SEO advice of the year came from the people least qualified to give it? That is the charge Charles Floate leveled when I sat him down for this episode of AI Rabbit Holes. Charles is an 18-year SEO veteran who scaled PressWhizz into one of the fastest-growing link-building platforms in the space, close to a million dollars a month, as its CMO. So when he says Google is gaslighting marketers, it carries weight. If you want to go deeper on building content systems that survive AI search, I put together a guide on how to automate SEO content with AI that ranks without tripping spam filters.
Why Google Is Gaslighting SEOs
You have probably seen this movie before. Every few years Google publishes guidance, and every few years the leaks tell a different story. Charles points to a clear pattern. The advice does not come from the engineers building the models. It comes from the same policy and documentation teams that have shaped Google's public messaging for two decades. When you compare that messaging against reality, DOJ testimony, and the document leaks, the gap is obvious. A lot of what lands in the official docs is factually incorrect versus how the systems actually behave. So why keep doing it? The incentive is money and resources. Counteracting aggressive SEO tactics at scale costs enormous processing power. Spreading a message through PR, conferences, and SEO influencers costs almost nothing. Right now Google would rather pour its compute into new model iterations than into policing techniques it can simply talk you out of using. There is a second layer with AI specifically. The guidance doubles as corporate presentation, a way to show investors and the digital PR world that Google is helpful and in control. How helpful the guide actually is stays up for debate.What Actually Works in AI Search
Almost no one can attribute AI citations accurately, and most do not even realize it. Attribution is the hardest problem in this new era, so start there. Unless someone clicks through from an AI answer, you get no direct attribution. When there is a click, server logs are the only reliable source. Everything else filters out data, misses referrals, and gives you numbers you cannot trust. If your brand gets cited without a click, the trail goes cold. You cannot cleanly connect a ChatGPT recommendation to a later branded search and a sale. The honest move is to track visibility growth across query clusters against your direct traffic and leads, then estimate. A simpler tactic works too. Add a required "How did you find us?" field in your checkout or lead form. It adds a little friction, but it captures attribution that no tool can. Here are the levers Charles says still move rankings in AI search:- Tokenization and chunking. Specific token structures around specific topic clusters get preferred by the models. This is fundamental to how they work, no matter what the guide says.
- Extractability. AI Overviews pull exact chunks of your content, summarize them, and cite the source. Structure your content so a model can lift a clean passage as-is.
- Co-occurrence and citations. None of this is new to SEO. The systems that adapted to these signals before will adapt again.
- Domain authority. Without it, nothing else matters. You can publish 100 posts tomorrow and watch zero of them index.
Parasite SEO Is the Real AI Search Hack
Most people racing to fine-tune their own site for AI search are wasting their time. That is a hard claim, and Charles backs it. Parasite SEO means you stop relying on your own domain and leech onto a stronger one instead. You publish a LinkedIn Pulse article or buy a newspaper editorial, and you inherit that domain's link authority, root authority, and topical authority. The signal boost is massive compared to your own zero-authority blog. High-authority domains also sit above the threshold for many anti-spam classifiers, so content that would nuke your own site can rank just fine on theirs. This is why Charles argues a strong YouTube channel plus owned distribution beats your own website in most cases. While everyone chases llms.txt and schema hacks, the math is simple. If your domain has no authority, those tricks change nothing. Post 100 blog posts on a dead domain and none of them rank. Post those same 100 as parasite pages on Medium or LinkedIn Pulse and many of them stick. That difference is raw domain power. YouTube, Reddit, and LinkedIn are top-tier sources that LLMs lean on constantly. If you understand how the retrieval process works, you stop fighting your own domain and start borrowing authority from theirs.Programmatic SEO vs Scaled Content Abuse
There is a line between programmatic SEO that wins and the scaled content abuse Google penalizes. Most operators sprint right past it. The failures usually share the same fingerprints. They spin up 400,000 pages, which is pure index bloat you will never get crawled. They skip internal linking, clean URL structures, schema, and unique imagery. They stuff meta titles until the pages read like templates, because they are. Worst of all, the pages are linear clones. "This city Italian restaurants" repeated across every location, ignoring how people in different places actually search, spell, and phrase queries. That is what most of us just call AI slop. The fix is a genuine unique angle. Charles builds client systems that pull from stored meeting context, podcast transcripts, and proprietary data so every page carries something the competition cannot scrape. Telling a model to recycle the top five Google results is a losing strategy unless you already rank like Forbes. If you want the non-negotiables for a content system that actually moves the needle, two stand out:- Positive and negative datasets. Most people never build them. Show the model a dozen great examples and a few hundred bad ones from the client's existing posts, and you fundamentally reshape its output through better RAG and fine-tuning.
- Real media over AI images. Charles swaps the image-generation step for a screenshot agent that captures real blog posts, testimonials, and reviews. Those unique images add relevance, freshness, and information gain that AI-generated art never will.