This past Friday, the warning came in Google's updated Search spam policies, whose definition of spam now includes techniques used to “manipulate generative AI responses in Google Search,” and Google says violations can cause pages or entire sites to rank lower or be omitted from results altogether.
For GEO/AEO companies built on high-volume blogs, shallow comparison content, and heuristic SEO fixes dressed up for AI search, this should be terrifying. For Second Wind, it's validation.
The old SEO playbook is on its way out
A lot of “GEO” today is run with outdated SEO tactics. You know the old playbook: publish more pages, write more “best X” content, add schema, track mentions, and reverse-engineer model outputs until the web is flooded with content designed to push systems toward a preferred answer. That approach was always fragile. Google just made the risk explicit. Attempts to manipulate generative AI responses now fall directly inside its definition of Search spam.
The industry term for this is “recommendation poisoning”: using biased listicles, hidden instructions, or distorted web signals to try to push AI systems toward a company.
This policy change is helpful for AI users. AI search does not behave like a traditional results page. A blue-link search page rewards visibility. AI search produces answers, comparisons, summaries, and recommendations. When a buyer asks which vendor to choose, the system has to figure out what the company does, who it serves, how it compares, what proof exists, and whether the sources are strong enough to support a recommendation.
More content, in a lot of cases, makes a company's evidence graph noisier, more repetitive, and harder to trust.
More content is not more credibility
High-volume content made sense when the job was to capture keywords. AI search is built on an entirely different foundation.
The systems need clear company facts, current information, source-backed claims, verifiable proof points, fair comparison context, buyer-fit explanations, and public references that support the company's positioning. That kind of information helps both humans and AI systems evaluate a company.
Thin content creates the opposite effect. It may create a short-term lift in mentions, but mentions are not the same as trust. So much of the GEO market is exposed here. Despite what SEO agencies may claim, a prompt-targeted blog post with weak information value is not infrastructure. A generic “best vendors” page with no real methodology crosses the bridge from useless to harmful.
Moreover, a dashboard showing that a brand got mentioned does not tell you whether the model understood the company correctly, trusted the right sources, and recommended it in the buyer context that actually drives selection and conversion.
What we should be asking is, what evidence is the system using to understand the company, and is that evidence accurate, complete, current, and easy to retrieve?
This is the bet we made
Our core thesis at Second Wind has been consistent: AI is a new buyer, not another search box. (We've written about that distinction before.)
A lot of the market is still trying to evaluate AI search with old SEO tools. Keyword rankings, generic visibility scores, and low-volume manual chatbot tests are not enough to understand whether a company is being accurately represented in buyer-facing AI conversations.
We look at raw model responses, competitive win rate, positioning alignment, citation behavior, AI referrals, and agent/crawler activity. We care whether the system understands the company, cites the right sources, frames it correctly, and routes high-intent users toward the right parts of the business. That requires a different product.
Second Wind builds dynamic reference infrastructure: public, crawlable, source-backed company information that helps AI systems and human buyers understand what a company does, where it fits, how it compares, and what evidence supports its claims. Instead of trying to cheat the system, we make the truth easier to find, verify, and use.
Fix the evidence, not just the output
The old SEO instinct is to optimize around surface signals. Second Wind starts with the underlying information environment.
We organize the information AI systems actually need: company positioning, services, buyer fit, proof points, comparison context, public sources, FAQs, source links, and areas where the company is commonly misunderstood.
Then we monitor how AI systems represent the company across real buyer prompts. When a model misses a core proof point, cites stale information, misunderstands the ICP, or frames the company poorly, the answer is rarely to pump out another blog post.
Sometimes the answer is a clearer source. Sometimes it is a better reference page. Sometimes it is structured data, a fair comparison page with real methodology, or a correction to inconsistent information across the web.
That work is far more durable than spinning up content at scale.
So sure, trust the SEO consultant telling you this is just another ranking game if you want. Just be ready when platforms start treating that playbook like spam.
Google is only the first signal
Google's update is about Search, but the larger shift is bigger than Google.
Buyers already use ChatGPT, Gemini, Claude, Perplexity, and Copilot to compare vendors, pressure-test options, and shorten research cycles. Over time, more of that work will move from human-assisted search to agent-mediated evaluation.
Those agents will not want 800-word blog posts padded for keywords. They will need structured, current, verifiable company information they can query, compare, and act on.
This is where agent-to-agent infrastructure becomes important.
Companies will need clean reference layers, evidence graphs, source-backed positioning, citation intelligence, buyer-context monitoring, and eventually agent-interactive endpoints that let AI systems evaluate a company directly instead of scraping a messy marketing site and guessing.
This is the direction Second Wind is building toward.
GEO is bifurcating into infrastructure and spam
Google's policy update strengthens the case for GEO. AI search is now important enough that platforms are policing attempts to manipulate it.
This validates the category while also applying pressure on its weakest links.
The GEO market is going to split. Many companies will keep selling old search tactics into a new interface: more content calendars, more prompt-targeted articles, more shallow “best X” pages, more heuristic recommendations, and more dashboards that confuse visibility with buyer impact.
Others will build the infrastructure for AI representation: structured reference layers, evidence-backed positioning, source hygiene, citation intelligence, buyer-context monitoring, and agent-readable company information.
Second Wind was built for that second path.
The companies that win AI search will make their accurate information easier to retrieve, evaluate, and trust. They will give AI systems a cleaner factual basis to work from.
In the age of agents, truth is the real optimization. This has been our thesis from the beginning.
This past Friday, Google validated it.