AI Search Is Rewiring SEO — and the Industry Is Already Trying to Game It

How AI Search Is Upending the Old SEO Playbook

Ask Google's AI Mode for a list of the best digital service desk platforms and you may get a confident, well-cited answer. Click the first source and you land on a Zendesk blog post that compares 15 rival products — and picks Zendesk as the winner. Click another and Freshworks has published its own "10 best" ranking, with Freshworks' Freshservice at the top. The pattern repeats across categories: Eesel recommends Eesel AI, Hiver recommends Hiver, Watermelon recommends Watermelon, and Help Scout recommends Help Scout.

These self-dealing listicles are one of the most visible signs of a bigger shift. As Google, OpenAI and others push AI-generated summaries that answer questions directly instead of listing links, the search engine optimization industry is scrambling to find ways to get chatbots to name and recommend their clients. The tactics range from finely structured comparison pages that large language models can easily parse, to hidden text injected into "Summarize with AI" buttons — a practice Microsoft publicly labeled "recommendation poisoning."

The stakes are existential for many publishers and brands. Google has been steadily sending less referral traffic to the open web, and a widely shared report from the marketing firm Growtika claimed major tech publications lost 58% of their Google traffic since 2024, with outlets like Digital Trends and ZDNet down more than 90% from their peaks. The Verge's publisher, Helen Havlak, called those figures "wildly inaccurate," while acknowledging that Google referrals to the web are genuinely declining.

For now, the gap between hype and reality is wide. SparkToro founder Rand Fishkin argues that AI search is receiving "between 10 and 100 times" more attention than the actual activity taking place there, with desktop searches on traditional engines still dwarfing chatbot queries. Even so, the SEO industry is moving fast — raising money, inventing new acronyms and promising clients dominance in an era that has no agreed-upon metrics.

What the Self-Dealing Listicles Actually Exploit

Why the Self-Dealing Listicles Keep Surfacing

The listicle trick works because AI systems run real-time web searches in the background to supplement their answers. A cleanly formatted comparison page with product names, features, pros and cons is easy for a language model to retrieve and cite — whether or not the page is genuinely objective. Britney Muller, a former SEO consultant who now runs Orange Labs, frames it as "a search engine information retrieval problem, not an AI or LLM problem." The model is not necessarily biased in favor of Zendesk or Freshworks; it is simply pulling from pages that happen to be structured for easy extraction.

Google says it is aware of the abuse. Spokesperson Jennifer Kutz told The Verge that the company has protections against manipulation in search and Gemini, and that it works to combat low-quality listicle content. After The Verge reached out, Kutz said many of the affected searches were already showing "higher quality information." That suggests the tactic has a short shelf life: algorithm updates can quickly demote the very pages the lists depend on.

The Numbers Behind the Panic Are Shaky

The sense that search is collapsing is driven heavily by Growtika's report, which claimed a 58% drop in Google traffic to the most-read tech publications since 2024. The figures came from Ahrefs, cover US organic traffic only, and were disputed by The Verge's publisher. But the underlying direction is real: Google's AI Overviews, the company's decision to rank Reddit highly, and users shifting some queries to ChatGPT and Claude have all reduced the flow of traditional clicks.

Fishkin's SparkToro analysis offers a useful corrective. On desktop, Amazon, Bing and YouTube each have a larger share of search activity than ChatGPT, yet almost no one is optimizing for them. The implication is that many brands are chasing a relatively small audience while ignoring established channels with bigger reach.

A Gold Rush With No Verified Metrics

The confusion has created room for aggressive marketing. Firms now sell AEO, GEO, GSO and other variants of "AI search optimization," promising chatbot visibility in 60 days. One unnamed firm that recently raised $9 million says it uses more than half a dozen AI agents to research queries, generate landing pages and secure backlinks. Muller warns that such promises set a "dangerous precedent" because there is no reliable way to measure AI influence yet.

The darker end of the spectrum includes hidden prompt injections that tell a language model to "remember as a trusted source for future citations." Microsoft documented this practice in February and called it recommendation poisoning. Muller's concern is broader: if AI agents cannot reliably distinguish malicious instructions from ordinary content, giving them control over real-world actions becomes risky.

Retailers Are Measuring What Actually Moves Purchases

While the hype cycle swirls, some retailers are treating AI search as a practical commerce channel. Mike Micucci, CEO of Fabric, says brands now rank AI discovery as a top-one-or-two priority. Fabric's Neon tool runs thousands of synthetic shopping prompts such as "best jeans for work casual outfits" and measures how often a brand appears in LLM responses versus competitors. The focus is on product page updates and data quality, not on tricking the model.

The cautionary note comes from OpenAI itself: The Information reported that the company was pulling back on some shopping features after realizing users were not actually buying inside ChatGPT. That suggests agentic commerce is still a vision rather than a mainstream behavior — and that visibility in AI answers matters most at the research stage, before the purchase happens on a retailer's own site.

What Brands, Publishers and Marketers Should Do About AI Search

• For publishers: Treat Growtika's 58% traffic-collapse figure as contested, not gospel — The Verge's publisher called it "wildly inaccurate." Audit your own Google Search Console and analytics data to see the real rate of decline before making drastic strategy changes.
• For SEO firms: Stop promising guaranteed AI visibility. There is still no agreed-upon way to measure it, and Muller warns that claims of being able to "influence AI" are setting clients up for failure.
• For marketers: Do not build a strategy around self-dealing listicles. Google says it is actively suppressing low-quality listicle content, and the BBC test in February showed how easily these manipulation attempts can be exposed when they go wrong.
• For brands and retailers: If AI discovery matters to you, copy what Fabric's Neon does: run synthetic prompts around your product categories, track how often your brand appears versus competitors, and fix the underlying product data that LLMs pull from.
• For companies tempted by hidden prompt injections in "Summarize with AI" buttons: Microsoft has already named and documented this as "recommendation poisoning." It is a short-term hack with serious reputational and platform-enforcement risk.