AI search optimization should begin by confirming that pages meet the conditions for indexing and snippets, rather than by adding “AI-specific” files. According to Google guidance relayed in Tenten’s August 2026 article, AI Overviews and AI Mode do not require special schema or additional machine-readable files; pages must still be indexed, allow snippets, and comply with existing search policies. This guidance applies only to the Google features described and cannot be generalized to all AI platforms. Eligibility also does not guarantee citation.
Check indexing eligibility before looking at whether answers use your content
Technical checks should distinguish between whether a page can be indexed by a search system and whether an AI answer uses its content. The first concerns eligibility; the second involves visibility and quality. Based on the Google guidance above, check each important page:
- Is the page indexed, rather than merely accessible?
- Does the page allow search systems to generate snippets?
- Does the page comply with existing search policies?
A single AI query cannot verify these conditions. Absence from one answer is not enough to establish that a page is ineligible for indexing, and appearing once does not prove that its content will continue to be used.
Schema must match the page content; it is not an additional requirement
Tenten’s August 2026 article notes that structured data can still help search engines understand a page, but it must match the visible text. Do not just check whether markup has been added. Check whether the information it describes actually appears on the page and conveys the same meaning.
For example, if markup describes an article’s content, each item should be checked against the visible text. Claims absent from the page should not be added merely to provide more machine-readable information. Helping systems understand content does not guarantee citation. The available material does not provide enough evidence to conclude that adding schema will necessarily increase visibility.
Evaluate llms.txt according to platform needs
A June 2026 Search Engine Journal report described llms.txt as a proposal rather than a standard. The report quoted John Mueller’s advice: invest time in creating the file when an AI platform that can bring you customers says it needs it.
This advice does not mean that no platform will ever need llms.txt, or that you must create it now. Before investing time, consider which platform is requesting it and whether that platform is relevant to your business. A comment on a proposal at that time should not be treated as a permanent rule.
Evaluate citations, accuracy, and conversions separately
Tenten recommends using a fixed set of queries with commercial intent, testing them repeatedly across multiple platforms at different times, and reviewing the results alongside website analytics. Track at least the following separately:
- Whether the brand or URL appears, and which pages are cited.
- Whether the cited content actually supports the answer and whether the answer contains any incorrect descriptions.
- Whether AI referral traffic and related conversions increase.
For example, if an answer mentions a brand but misstates the scope of its services, it should not simply be counted as a visibility success. A single screenshot captures only one result. Changes in traffic and conversions must also be examined separately and cannot be attributed directly to a particular technical change.
AI search optimization: perspectives from across the web
How can you check whether a website is blocking crawlers?
Tenten recommends checking robots.txt to see whether it inadvertently blocks important crawlers from platforms you want to reach. Checks should focus on the target platform and its relevant crawlers, rather than assuming that all AI platforms use the same crawlers. This is a step toward diagnosing access barriers; unblocking a crawler does not guarantee indexing or citation.
Does important content have to use server-side rendering?
There is no blanket answer; check the target platform and the actual page. Tenten recommends checking whether important content uses server-side rendering rather than relying on JavaScript to appear, as a way to assess whether it can be crawled. However, the available material does not establish that all platforms require server-side rendering or that pages using JavaScript are always unreadable.
How can you distinguish AI referral traffic in website analytics?
Tenten recommends setting up a custom channel group in GA4 to monitor traffic from sources such as ChatGPT, Perplexity, and Gemini separately, while using Search Console to review search and indexing status. If you also need advertising or B2B audience analysis, you can consider Meta Pixel or the LinkedIn Insight Tag; they are not requirements for every website.
References
- Google Says LLMs.txt Is Purely Speculative… For Nowwww.searchenginejournal.com · 2026-06-02T10:08:07+00:00 — Source used in the article above.
- GEO 正在贏得縮寫戰,但還不是 AI 搜尋的共同語言tenten.co · 2026-08-20T18:53:30.000Z — Source used in the article above.
- 流量暴增秘辛!5 步驟破解 Perplexity GEO,讓 AI 狂推你的網站tenten.co · 2026-04-22T05:43:24.000Z — Discusses: How can you check whether a website is blocking crawlers?; Does important content have to use server-side rendering?

