AI search is not killing traditional search, purely AI-generated content can still rank, and llms.txt files do little to improve visibility in AI answers. Those are among the findings of a new data-driven review that challenges seven widely repeated claims about how artificial intelligence is reshaping search.
The review, published by MarTech, argues that many of the most persistent claims about AI and search follow a familiar pattern: they sound plausible, get repeated across social media, and harden into conventional wisdom without ever being checked against real data. The result, according to the analysis, is a stream of misinformation that often serves as a scare tactic to generate sales.
The first myth holds that AI platforms such as ChatGPT, Claude, and Perplexity are cannibalizing traditional search. The data tells a different story. The State of Search Q2 2026 report found that AI search and traditional search are growing at roughly the same rate quarter over quarter. Rather than one replacing the other, the two are expanding together, with AI adding a new layer on top of existing search demand.
There is a kernel of truth in the concern, however. Clicks to websites are falling as Google increasingly answers queries directly without sending users anywhere. Clicks to non-Google-owned desktop results are at their lowest level since April 2025, even as overall search demand continues to grow.
A second myth claims that organic traffic cannot grow in a zero-click environment dominated by AI Overviews and AI Mode. The review points to a local business that recorded its highest month of organic website clicks in July 2026, driven entirely by nonbranded efforts across blog content and service pages. The analysis cautions that local businesses fall into a different risk category than the broader search landscape, and not every site will see that kind of growth. Still, the blanket claim that organic wins are impossible does not hold up.
The third myth is that purely AI-generated content cannot rank. In a test, four articles generated entirely through AI with careful prompting and light human review, published between March and September of last year, are still ranking and performing. The review acknowledges that much AI-generated content is low-quality slop, and that Google has improved at recognizing unoriginal, keyword-stuffed output. But the distinction that matters is process, not whether a model wrote the copy. Content that starts with a good idea, genuine expertise, original data, or a specific point of view can perform well regardless of how it was produced.
On the technical side, the review debunks the claim that adding an llms.txt file meaningfully improves AI visibility. Google has said explicitly that it does not use llms.txt for AI search discovery. A study that tracked adoption across 10 sites for 90 days before and after implementation found no measurable change in AI crawl frequency or AI-referred traffic for most sites. The review concludes that llms.txt is infrastructure, not strategy: it cannot hurt to have one, but it will not move the needle.
Another technical myth holds that modern AI crawlers can render JavaScript the way a browser does. Profound's study of AI crawler behavior found that emerging AI search providers lack JavaScript rendering capabilities, making server-side rendering crucial for AI visibility. The myth persists partly because large language models sometimes misrepresent their own abilities. In one example, ChatGPT claimed it could summarize an article whose content was hidden behind client-side JavaScript, recited key points, and then admitted it had not been entirely truthful.
Together, the findings suggest that search professionals should verify claims against data before allowing them to shape strategy. AI is changing search, but the evidence indicates it is adding to traditional search rather than replacing it, and that value, not the identity of the author, remains the deciding factor in whether content ranks.