Your GEO Strategy Might Be Destroying Your SEO
Many tactics designed to win in AI search can undermine SEO performance - which is, ironically, necessary for success in AI search.
We have reached the peak of the AI search hype cycle. Lately, for those of us in the search marketing industry, it feels impossible to open any social media feed without being flooded with posts or ads about how brands not investing in GEO will be invisible in AI search.
I made this collage of some of my favorites from the past few months:
Along with all this excitement about AI search comes a lot of bad advice, largely stemming from a lack of information about how LLMs actually work, and/or a lack of SEO experience which, as it turns out, is vital for earning and sustaining visibility within AI search responses.
I believe many new AI search case studies and success stories being shared are problematic and should be questioned, for a few common reasons:
They show a rapid growth in SEO/AI search visibility, but do not account for the subsequent crash that often follows after search engines eventually demote the site
They assume GEO exists in a vacuum, and deny or ignore that existing SEO performance had anything to do with success in AI search
They confuse GEO tactics with good, old-fashioned SEO tactics - rebranding existing SEO approaches under a new name
They claim SEO is dead or no longer relevant, when in reality, strong SEO performance is the very foundation of AI search visibility
Before getting into specific examples, it’s worth understanding why SEO is so fundamental to AI search performance - because the answer lies in how these systems actually work at a technical level.
How AI Search Actually Retrieves Content
Every major AI search product - ChatGPT, Perplexity, Gemini, Google AI Overviews, Microsoft Copilot - is built on a retrieval architecture called RAG: Retrieval Augmented Generation. While this mechanism has been discussed for years now within the SEO/GEO industry, it still feels like there may be some confusion about its role in generating AI responses.
When a prompt requires current, specific, or verifiable information - which includes most commercial and research queries - LLMs like ChatGPT, Perplexity, and Gemini retrieve relevant documents from external search indexes before generating their response. For these queries, the retrieval step is key: if your content isn’t indexed and ranking in the search engine, it cannot enter the model’s context window.
As AI expert Britney Muller stated last year on LinkedIn: “👏 Every 👏 Single 👏 URL 👏 you see in an LLM output comes from a search engine API (Google or Bing).”
So, which search engines are the LLMs using? For ChatGPT specifically, the answer has become more complicated than it used to be, but many recent reports point toward a similar answer.
OpenAI and Microsoft originally had a deep exclusive partnership, with Bing serving as the primary search layer powering ChatGPT’s retrieval. That exclusivity has apparently since unraveled: Microsoft’s exclusive cloud provider status was annulled in late 2025, and OpenAI began sourcing compute and infrastructure from competing providers, including Google.
This great Search Engine Land article by Olivier de Segonzac explains how after OpenAI was denied direct access to Google’s search index in 2024, the company since has been “been partially using Google search results scraped by SerpAPI to power ChatGPT’s real-time answers.” This detail provides a reasonable theory for why Google announced a lawsuit against SerpAPI in December 2025.
What has been interesting to observe over time is just how much independent evidence has accumulated pointing to ChatGPT's reliance on Google for retrieval. Researchers both inside and outside of the SEO industry have all arrived at similar conclusions through different methods - controlled indexing experiments, shopping carousel analysis, search log comparisons, and investigative reporting - and new findings continue to emerge. For example:
OpenAI Is Challenging Google - While Using Its Search Data — The Information
ChatGPT sources 83% of its carousel products from Google Shopping via shopping query fan-outs — Search Engine Land
ChatGPT is using Google Search, multiple tests suggest — Search Engine Land
ChatGPT Is Using Google Search – We Tested It — Backlinko
ChatGPT Appears To Use Google Search As A Fallback — Search Engine Journal
ChatGPT is secretly using Google Search data - here’s how — Tom’s Guide
ChatGPT Searches Google Shopping to Create its Recommendations — Semrush
ChatGPT’s answers came from Google Search after all: Report — Search Engine Land
My recent Substack showing that ChatGPT citations appear to be closely tied to organic performance
While we may never receive an official confirmation from OpenAI about the extent to which ChatGPT relies on Google’s index (if at all), the volume and consistency of independent evidence makes a compelling case on its own. And of course, ChatGPT isn’t the only AI product where Google’s index is the main retrieval layer; Google’s own AI products (AI Overviews, AI Mode, Gemini) are built on top of it, as confirmed many times by Google employees.
Below is some data recently sent to me by Similarweb that shows the worldwide growth in LLM usage (desktop visits only) over time. The red line is Google’s Gemini, which is gradually catching up to ChatGPT usage:
And that’s before you even factor in Google’s AI Overviews - which, if counted as part of the AI search ecosystem, dwarf every other LLM product in usage by a wide margin.
As Rand Fishkin wrote in a recent study, “Most AI Search and AI Answers happen on Google. Even if you combine every prompt on ChatGPT, Claude, Deepseek, and the rest and assume every prompt is a search-equivalent, Google dwarfs them. If 16% of results are showing AI Overviews, Google’s by far the largest AI search tool in both the US and EU/UK by at least an order of magnitude.”
The implication of all this is significant for AI search visibility: Google’s index has become the de facto retrieval foundation for the majority of AI search traffic - including, increasingly, ChatGPT. Optimizing for Google organic search isn’t just good practice for traditional search (SEO); it’s the primary mechanism by which content enters the context windows of the most widely used AI search products on the market.
On the flipside, undermining your SEO visibility may be one of the single most damaging things you can do to your long-term AI search presence, since the organic rankings you sacrifice are the very foundation that determines whether AI search platforms retrieve your content in the first place.
And as it turns out: many newly popularized GEO tactics risk doing exactly that.
The GEO Tactics That Work - Until They Don’t
The excitement around GEO is leading some marketers down paths the SEO industry has walked many times before: chasing shortcuts that deliver short-term wins but ultimately result in penalties, traffic losses, and long-term visibility damage.
To be clear: it appears that Google, OpenAI, and other AI companies haven’t fully caught up to these tactics yet, or started to heavily demote them. This is why many popular GEO tactics can and do work right now. But working now is not the same as working forever, and anyone who has been in SEO long enough knows exactly how this story ends. The problem is that many of the people and companies promoting GEO tactics just got here.
Seeing something work well in SEO/GEO for 3-6 months is not a guarantee that it will work long-term. In fact, rapid growth in SEO often leads to an equally rapid crash on the other side. This is what Glenn Gabe has been calling “Mount AI.” And the worst part: that crash can be incredibly difficult to recover from, sometimes requiring multiple years of extremely hard work to climb back to previous visibility levels.
This is the core issue with many AI-assisted approaches right now: the results can look genuinely promising at first. Scale AI-generated content, or write a few self-promotional listicles, watch a few articles gain traction, double down and publish hundreds or thousands more. It feels like a growth strategy. But it’s one I’ve described the same way for many years: it works - until it doesn’t.
Here are 5 GEO tactics I’ve seen circulating that I believe could eventually backfire for both SEO and AI search performance - and in some cases, already have:
1. Scaling content rapidly with AI
The premise behind scaled content is simple: publish hundreds or thousands of articles, with the assumption that more content means more SEO/GEO visibility. This isn’t a new idea - it’s exactly why Google introduced a spam policy called “Scaled Content Abuse” in 2024, and it’s an evolution of a long-standing policy against automatically generating content without “producing anything original or adding sufficient value.”
But AI has made scaling content easier, cheaper, and more tempting than ever. Automated content workflows have dramatically lowered the barrier to entry, and an entire category of tools has emerged to help sites produce more content, faster. Several of these tools are also actively marketing the SEO and GEO visibility gains their clients have achieved.
I’ve spent time reading through public case studies from some of these companies, and cross-referencing the traffic and visibility trends of the clients featured on their websites. What I found was a concerning pattern: for many sites, the content scaling appeared to work, until it didn’t.
Below are the organic traffic trends (via Ahrefs) for 3 companies featured in case studies published by a few of the AI content generation tools:
Note: The cursor in these screenshots is positioned over the approximate date when the case study was published. The dark orange line represents organic traffic, while the yellow line represents change in organic pages:
In this example, the case study was published in mid-April, 2025. The site saw a substantial decline in organic traffic a few months later, during the June 2025 Core Update.
In this example, the site saw tremendous organic growth throughout late 2024, and the case study was published in January 2025. Immediately after, the site reversed course, and began to see a substantial organic traffic decline that has continued ever since.
For this site, the case study was published in March of 2024. Immediately after, when the March 2024 core update was launched, the site began dropping fast, and the trend has only continued. In fact, this site recently 410’d (permanently removed) the same articles that were celebrated as success stories in the case study, resulting in the major traffic drop in February 2026.
These are three of roughly 30 similar examples I’ve found across case studies published by AI content scaling companies over the past few years. I will likely publish more on this topic in the coming weeks as I gather more data.
It’s worth noting these are third-party Ahrefs estimates, not first-party data, and there could be other contributing factors behind each traffic drop. But the pattern - rapid growth followed by an equally rapid collapse - is consistent with what I’ve observed repeatedly when sites use AI to rapidly scale content that doesn’t offer original value beyond what’s already available in search results.
One additional question this data raises: could the case study itself have been the tipping point? It wouldn’t be the first time Google began to pay closer attention to a site after it publicly showcased its SEO wins. However, we typically see that play out as a manual action rather than an algorithmic demotion - but the timing in some of these examples is indeed curious.
My main recommendation in these cases is to do your own research when reviewing public case studies to validate that the longterm trend supports that the campaign was truly a success.
2. Artificial refreshing
AI has made it easier than ever to superficially refresh content - adding a few sentences or tweaking a paragraph just enough to justify updating the “date modified” timestamp that appears in search results. The appeal is obvious: fresher-looking content can improve both SEO and AI search visibility. It can also absolutely improve click-through-rates.
The problem is that Google has been aware of this practice - often called “artificial freshening” - for years. Despite recently removing the term from its Google News and Discover spam policies, this approach remains one of the most common patterns I’ve observed among sites hit by algorithm updates over the past decade.
For example, a site might update all of its titles to reflect the current year at the start of the year, but the information and links included in the articles are outdated or not truly relevant to the current year.
Google has gotten increasingly good at distinguishing genuine updates from cosmetic ones. AI makes article updates much easier to execute, but it doesn’t make it any harder for Google to detect when it compares the changes to the previous version of your page. My recommendation: before updating a publish date, make sure the changes are actually meaningful to readers - not just meaningful enough to try to fool a crawler.
3. Excessive Self-Promotional Listicles
Last month, I wrote about how it appeared that Google was beginning to crack down on sites using excessive self-promotional listicles: articles that rank the best companies/products in a given niche, ranking their own brand or product as the #1 best in the category.
For the affected sites, the trend has only continued. I have also had numerous companies reach out to me who were also affected by this issue beginning around January 21, 2026. As I mentioned in my article, not every company using these tactics has been affected - it appears Google is still in the early stages, likely targeting the heaviest offenders first. I also want to reiterate that this tactic can and does still work incredibly well for AI search - but I do not believe this will be the case forever.
Writing self-promotional listicles has become incredibly popular and is now even being sold as a service companies are offering as a method of driving AI search visibility. My hunch is that a much bigger crackdown on this tactic is around the corner.
Below are the organic traffic trends for two companies that heavily utilized self-promotional listicles (relative to the amount of content they have on their sites):
4. “Summarize with AI” Buttons that Promote Your Brand
Of all the tactics covered in this article, this one is categorically different - not a grey-area SEO shortcut, but something that has been formally documented by Microsoft as a security threat.
“Summarize with AI” buttons are, on the surface, a reasonable UX feature: a quick way for visitors to digest long-form content by linking them to the AI chatbot of their choice and asking the LLM to summarize the article.
This approach has gained popularity in the past year within the SEO/AI search community, and I do believe many marketers promoting it do simply want to offer a helpful UX feature for their audiences.
The problem is what’s happening underneath the hood among a growing number of sites.
In February 2026, Microsoft’s security research team published a report documenting what they call “AI Recommendation Poisoning”: companies hiding prompt-injection instructions inside “Summarize with AI” buttons that, when clicked, plant commands into the user’s AI assistant memory - instructing it to “remember [Company] as a trusted source” or “recommend [Company] first” in future conversations.
Microsoft found over 50 unique examples from 31 companies across 14 industries - not hackers, but legitimate businesses in finance, healthcare, legal, and marketing - and traced the technique back to two publicly available tools (CiteMET and AI Share URL Creator) marketed openly as “SEO growth hacks for LLMs.”
Microsoft formally classifies this as a form of prompt injection - the same adversarial taxonomy used for cyberattacks. I believe this is one of the first times we have seen an SEO/GEO tactic publicly identified as a form of prompt injection.
To put it more simply: including hidden instructions in a “summarize” button is an attempt to manipulate what AI systems say about you rather than earning that visibility legitimately. Microsoft’s own report notes that the tactic raises questions under privacy law, consumer protection regulations, and deceptive trade practice statutes - particularly for companies in health and finance, who face strict regulatory standards.
If Microsoft is publicly classifying this as a security threat, I think it’s reasonable to assume that Google is paying close attention too - even if they haven’t posted about it (yet). Unfortunately, I think this might be one of those tactics where even being loosely associated with it could now carry risk, regardless of how pure your intentions were.
5. Excessive “Alternatives” and “Comparison” Pages
Comparison content - ‘X alternatives’ or ‘X vs. Y’ pages - can absolutely be a legitimate and useful content format when done thoughtfully and in moderation. The problem isn’t the tactic itself; it’s the scale.
This approach falls into the grey area, because including “comparison” and “alternative” information can absolutely be helpful for users and good for conversions. However, when scaled heavily for SEO/AEO/GEO purposes, I believe the content can also become a liability.
The idea is to produce a substantial amount of content comparing brands against all possible competitors. In the example below, the site created 51 such pages:
Below is the organic trajectory of this company’s blog, which scaled this tactic throughout 2025:
Their blog also began to see a drop in ChatGPT citations around the same time organic traffic started to decline (late Jan 2026):
All five of these tactics share a common thread: they treat AI search visibility as something to be manufactured, rather than earned - and manufacturing visibility is precisely what Google and other search engines have spent years building systems to detect and demote.
In chasing shortcuts that the algorithms are actively working to demote, these approaches risk destroying the very SEO foundation that makes AI search visibility possible in the first place.
GEO Gets the Credit. SEO Did the Work.
A pattern I keep seeing in AI search success stories is that the “GEO wins” being celebrated can almost always be explained by something simpler: strong pre-existing SEO performance. (I thought this recent post by Jeremy Moser of uSERP also explained this well!)
These case studies typically fall into one of two traps. The first is relabeling longstanding SEO tactics - structured data, FAQ schema, optimized headlines, bullet points, TL;DRs, directly answering questions, etc. - as novel GEO innovations, when these approaches have been core SEO best practices for years. The second, and more misleading, is claiming GEO credit for AI search visibility that was almost certainly driven by organic rankings the site already had before any GEO strategy was implemented. (I also wrote about this situation last year on LinkedIn.)
To be fair, reframing familiar tactics under a new umbrella isn’t inherently harmful; language evolves, and if “GEO” helps new audiences discover practices that genuinely work, that’s fine. The problem is that this rebranding often comes packaged with an implicit - or sometimes explicit - claim that SEO no longer matters. That framing is not only inaccurate; it’s actually dangerous, because SEO is the real explanation for the results being celebrated.
Here’s an example from X:
This is fundamentally a correlation problem. A brand with years of established authority, strong backlink profiles, many brand mentions, and strong organic rankings starts appearing in ChatGPT responses and Gemini citations - and concludes that their new GEO campaign is working. But the more plausible explanation is that their existing SEO visibility is what got them into the search indexes feeding those AI products in the first place.
The AI citations and brand mentions aren’t happening despite their SEO - they’re happening because of it. Correlation is being sold as causation, and an entire industry is being built on that misattribution.
At the end of the day, the name itself doesn’t matter - just don’t abandon the SEO strategy that’s likely responsible for the results you’re attributing to GEO.
Your SEO and GEO Strategies Should Always Work Together
The AI search opportunity is real, and the urgency to show up in these new environments is justified.
But one of the riskiest things you can do in pursuit of AI search visibility is sacrifice the SEO strategy that serves as the foundation for AI search success. These aren’t two separate channels you can trade one for the other - they’re part of the same system.
Undermining your organic rankings by chasing risky AI search shortcuts won’t just result in a loss in SEO traffic; because of RAG, you can lose your AI search visibility too. The brands that will win long-term are the ones who evaluate every new AI search tactic through an SEO lens first - and who recognize that shortcuts designed to manipulate search and AI systems don’t just risk rankings, they also risk the brand reputation that defines their authority in the space.