AI Search Myths Marketers Still Believe

Many AI search beliefs don't hold up against what engines and research actually say. SEO isn't dead, llms.txt isn't required, schema doesn't guarantee citations, one prompt doesn't measure visibility, and keyword stuffing doesn't work in AI answers. This guide checks ten common myths against primary sources and gives the better practice for each.

Myth 1: "SEO is dead"

Reality: Google says SEO best practices remain relevant for AI Overviews and AI Mode, and a page must be indexed and eligible for a snippet to appear as a supporting link (Google). AI search changes what you optimise for, not whether foundations matter. See Is SEO dead?.

Myth 2: "You need special markup to appear in AI Overviews"

Reality: Google says there are no additional requirements or special optimisations needed to appear in AI Overviews or AI Mode (Google). Structured data helps Google understand content (Google), but it's not a ticket into AI answers.

Myth 3: "llms.txt guarantees AI engines will use your content"

Reality: llms.txt is a proposal to help language models use a website at inference time (llmstxt.org). Google's AI features documentation lists no such requirement (Google). It can be a cheap addition, but treat it as optional and unproven.

Myth 4: "Keyword stuffing works for AI"

Reality: In the GEO study, keyword stuffing gave little or no improvement in visibility in generated answers, and on Perplexity it did about 10% worse than the baseline. Adding citations, quotations and statistics performed much better (arXiv). Google's spam policies also prohibit keyword stuffing (Google).

Myth 5: "One prompt tells you if you're visible"

Reality: AI answers vary between runs. Measure rates over repeated runs, with ranges (NIST). See Why AI answers differ every time.

Myth 6: "Blocking AI crawlers protects you without cost"

Reality: Crawlers have different purposes. OpenAI separates OAI-SearchBot, used for ChatGPT search, from GPTBot, used for training (OpenAI). Blocking a search crawler can remove you from those answers. Decide per crawler.

Myth 7: "AI just uses training data, so new content doesn't matter"

Reality: Many engines retrieve live web results. Google's AI features use a query fan-out technique, running multiple related searches (Google). Retrieval-augmented generation, described in 2020 research, combines a model with a retrieved document index (arXiv). New, indexed content can appear quickly.

Myth 8: "AI-generated content is penalised"

Reality: Google says it rewards high-quality content however it's produced; using automation primarily to manipulate rankings breaks its spam policies (Google). The test is helpfulness, not the tool.

Myth 9: "If you rank #1, you'll be in the AI answer"

Reality: AI answers often cite pages beyond the top result. In the GEO study, lower-ranked pages gained most from optimisation; for example, pages at rank 5 gained 115.1% from citing sources, while rank 1 pages lost 30.3% (arXiv). Ranking helps, but it isn't the whole picture.

Myth 10: "You can pay to be included"

Reality: Google says it doesn't accept payment to crawl or rank sites (Google). Be wary of vendors who promise guaranteed placement in organic AI answers.

Summary table

Why do these myths spread?

• Early experiments with small samples get treated as rules.
• Vendors sometimes sell certainty that doesn't exist.
• Engines change fast; advice from a year ago can be outdated.
• AI answers vary, so anecdotes seem to "prove" opposing claims.

How do you test a claim yourself?

1. Find the primary source: the engine's documentation or the original paper.
2. Check the date; engines change.
3. Run your own test with repeated runs and a control.
4. Look for a sustained change beyond normal variation.

What should you do instead of chasing myths?

Focus on the three things the evidence consistently supports: be crawlable and indexed, publish clear and well-sourced answers, and measure your visibility with repeated runs. Everything else is an experiment to test, not a rule to follow.

Checklist

• [ ] Primary sources checked before acting on advice
• [ ] Crawler decisions made per purpose
• [ ] Visibility measured as rates, not single answers
• [ ] Content improved with citations, quotes and statistics
• [ ] Promises of guaranteed AI placement treated with suspicion

Sources

• Google Search Central, AI features and your website
• Google Search Central, How Search works
• Google Search Central, Spam policies
• Google Search Central, Google Search's guidance about AI-generated content
• Aggarwal et al., GEO: Generative Engine Optimization, KDD 2024
• Lewis et al., Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks, 2020
• OpenAI, Overview of OpenAI crawlers
• llms.txt proposal