Optimizing for AI search is still SEO.
Confirmed
Google's guide says so. It speaks only for Google.
Google, AI optimization guide (opens in a new tab)
AI Overviews need extra files or markup.
Not required
Google: no new machine-readable files, AI text files, markup or Markdown.
Google, AI optimization guide (opens in a new tab) source 1Google, AI features (opens in a new tab) source 2
llms.txt lifts citations.
Not required
Google ignores it. SE Ranking found no effect across nearly 300,000 domains. That is correlation, not proof of zero.
Google, AI optimization guide (opens in a new tab) source 1SE Ranking llms.txt study (opens in a new tab) source 2
Allow OAI-SearchBot to appear in ChatGPT search.
Confirmed
OpenAI says each bot is independent. GPTBot is the training crawler.
OpenAI, crawlers (opens in a new tab)
Pages are cited when the engine's own search retrieves them.
Reported
Preprint, 158,099 answers: retrieval explains about 32% of variation, page properties 2.8%. Commercial questions, ChatGPT focus.
arXiv preprint, Oct 2026 (opens in a new tab)
Sources, quotes and statistics raise visibility.
Reported
Gains of 30 to 40% in a 2024 benchmark on older models. Treat as direction, not a forecast.
GEO paper, KDD 2024 (opens in a new tab)
The same prompt gives the same brands.
Reported
Under 1 in 100 repeats. Appearance rate over 60 to 100 runs is steadier than position.
SparkToro study, via SEJ (opens in a new tab)
Answering in the first two sentences gets you quoted.
Hypothesis
Common advice, easy to do. I found no controlled public test, so measure it on your own pages.
No public source
AI is already a large traffic channel.
Reported
0.32% of site traffic in one study, 1.08% in another on 2025 data. Small and growing.
SE Ranking traffic study (opens in a new tab) source 1Conductor benchmarks (opens in a new tab) source 2
A citation count is a ranking.
Confirmed
Bing: counts reflect how often pages are cited, not importance, ranking or placement.
Bing, AI Performance (opens in a new tab)