# What Technical SEO for AI Search Changes, and What Actually Affects Citations > Technical SEO for AI search comes down to access, readable HTML and accurate facts. What's documented, what's speculation, and how to check your SaaS site. Source: https://comppound.com/blog/technical-seo-for-ai-search-what-affects-whether-ai-cites-your-saas-pages (published 2026-10-30) Technical SEO for AI search is the work of making sure AI systems can reach your pages and read them. Most of it is existing technical SEO: allow the right crawlers, put content in the initial HTML, return correct status codes and keep your facts consistent. Be wary of claims about special files or markup that get you cited. [Google says outright](https://developers.google.com/search/docs/appearance/ai-features) that you don't need new machine-readable files, AI text files or special schema to appear in its AI features. Good technical work removes blockers. It won't by itself make an AI system recommend you, since that depends on your content and your reputation. But a page that can't be fetched can't be cited at all, so it's the right place to start. ## Separate what's documented from what's guesswork This field is full of confident claims with no evidence behind them. A simple division helps. **Documented by the companies themselves:** - Google's AI Overviews and AI Mode can use pages that are indexed and eligible for a snippet, with no additional technical requirements. Snippet controls such as `nosnippet` and `max-snippet` limit what's shown. - OpenAI's [crawler documentation](https://developers.openai.com/docs/bots) says OAI-SearchBot surfaces sites in ChatGPT's search features, and that GPTBot is a separate agent related to model training. - Crawlers respect robots.txt, with exceptions for some user-initiated fetches, as OpenAI notes for ChatGPT-User. **Reasonable inference, not confirmed:** - Crawlers that don't run JavaScript may miss client-rendered content. The safe response is the same whichever crawler it is. - Clear page structure makes content easier to extract. **Speculation, or contradicted:** - That any particular schema type causes citations - That an `llms.txt` file drives visibility. See [llms.txt for SaaS websites](/blog/llms-txt-for-saas-websites-what-to-include), and note that Google says it doesn't require such files for its AI features. - That a "citation optimization score" from a tool predicts anything ## Check access first: robots.txt and the firewall Say you run a 40-person SaaS for contract review. Check two things, as they fail independently. **Robots.txt.** Look for rules that name AI crawlers, plus any broad rules. A choice to block one agent doesn't have to apply to all of them. For instance, you can allow OAI-SearchBot for search while disallowing GPTBot for training: ``` User-agent: OAI-SearchBot Allow: / User-agent: GPTBot Disallow: / ``` Each company documents its own agents, and the names change, so check the current list before making changes. [Should SaaS sites block AI crawlers?](/blog/should-saas-sites-block-ai-crawlers-in-robots-txt) goes through the trade-offs. Note also that for Google, the `Google-Extended` token relates to AI training and some other Google systems, and it's not how you control appearing in Search or its AI features there. **Firewall and bot protection.** This one catches many companies out. Your robots.txt might allow a crawler while your CDN returns a challenge page or a 403. Some providers have added settings that block AI bots by default, or with a single click, so someone may have switched this on without thinking about the effect on visibility. Check your provider's settings and logs. OpenAI publishes IP ranges for its crawlers so that you can allow them. Your server logs can confirm this. Filter by user agent, check the response codes, and verify the IPs where you can, because user agents are easily spoofed. [Crawlability problems on SaaS sites](/blog/crawlability-problems-on-saas-sites-and-how-to-fix-them) has more on this. ## Put content in the initial HTML If your pricing table or feature list only appears after JavaScript has run, you're relying on every crawler to execute it. Google does render pages. You shouldn't assume the same of everyone else. Test it with `curl -s https://example.com/pricing`, then search the output for a sentence from the page. If it's missing, fix that page first, as described in [JavaScript SEO for React and Next.js](/blog/javascript-seo-for-saas-sites-built-on-react-and-nextjs). The contract review company's most important facts, such as plans, supported file types and integrations, should all be in that response. Also keep key facts in text. Information in images, videos or canvas elements is harder to use. Google's guidance mentions keeping important content in text form. If your comparison is only available as a screenshot, write it out in a real table too. ## Control what can be used You've got options if there's content you don't want shown. For Google, `nosnippet`, `data-nosnippet` and `max-snippet` all limit what's displayed. `nosnippet` prevents [content from being used as direct input for AI Overviews and AI Mode](https://developers.google.com/search/docs/crawling-indexing/robots-meta-tag), and it also removes regular snippets, so consider this carefully. `noindex` keeps pages out of Search altogether. Our [guide to meta robots](/blog/meta-robots-x-robots-tag-and-noindex-on-saas-sites) has the syntax. Check that none of these are switched on by accident. A `nosnippet` in a template can make a whole section ineligible for AI features, and nobody would notice it from looking at the page. ## Keep structure clean, without overdoing it Plain, semantic HTML helps any parser. Use one clear H1, headings that describe their sections, real lists and real `` elements. Let each section open with a statement that stands by itself, as in [writing content that survives AI summaries](/blog/writing-saas-content-that-survives-ai-summaries). Structured data is worth doing, for accurate descriptions and rich results, but it isn't a lever for AI citations. Google says that structured data isn't required for its AI features, while asking that it matches the visible text. Use Organization, Article and BreadcrumbList where they apply, and SoftwareApplication with care. See [does FAQ schema help in AI search?](/blog/does-faq-schema-help-in-ai-search) ## Make facts consistent and current For AI systems, this is arguably a technical issue too. If your pricing page, an old blog post and a docs page all say something different, there's nothing to tell a system which one is right. Practical steps for the contract review SaaS: - **Find contradictions.** Search your own site for old prices, retired plan names and former product names. - **Show dates honestly.** Visible "last updated" dates and accurate `lastmod` values in your sitemap help systems judge freshness. Changing dates without changing the content is misleading. - **Remove or update retired pages.** A page describing a discontinued feature should be redirected or clearly marked. - **Keep your company description consistent** across your site, directories and profiles. See [entity SEO for SaaS brands](/blog/entity-seo-for-saas-brands). - **Monitor mistakes.** When an answer gets something wrong, trace it to its source. See [fixing AI search brand misinformation](/blog/fix-ai-search-brand-misinformation-for-your-saas-product). ## Speed and reliability still count Live fetches may have short timeouts, though the exact limits aren't generally published. A page that takes many seconds to generate is at risk, whoever's asking. Static generation and caching solve most of this for marketing pages. Return proper status codes, as a friendly error page sent with a 200 may be read as content. ## Help new content get discovered Google's systems rely on its index, so the [normal indexing process](/blog/how-to-get-a-saas-site-indexed-by-google) applies. For Bing and others that support it, IndexNow lets you notify search engines when URLs change. It's simple to add, particularly on a site that updates often. See [Bing SEO for SaaS](/blog/is-bing-seo-worth-it-for-a-saas-company). ## A one-hour technical check 1. Read robots.txt for all AI agents and broad rules. 2. Check your CDN or firewall settings for bot blocking. 3. Filter a week of logs by AI user agents, and check the response codes. 4. Curl your pricing, top feature and top comparison pages. Is the content there? 5. Check those same pages for `nosnippet` and `noindex`. 6. Search your site for outdated prices and product names. 7. Run ten buyer prompts, and record what's cited. Does anything point to a page you'd rather they didn't use? Then measure with a monthly [AI citation audit](/blog/ai-citation-audit-for-saas-blogs-step-by-step-guide). Don't expect the technical fixes to deliver results alone. Going back to the contract review SaaS, unblocking a crawler is what makes the rest possible, but being recommended depends on having useful pages, and on [other sites talking about you](/blog/brand-mentions-and-llm-visibility-for-saas). To have this covered as part of ongoing work, [see how we approach SEO and AI search](/services/seo-growth). ## Frequently asked questions ### Do I need special markup to appear in AI answers? For Google's AI features, no. Google says there are no additional technical requirements beyond being indexed and eligible for a snippet, and that special schema or AI text files aren't needed. Other tools don't publish a markup requirement either. ### Does an llms.txt file help AI citations? There's little evidence that it does. Google says it doesn't need AI text files for its AI features. It's cheap to add if you'd like to, but don't expect it to drive visibility. ### Can AI crawlers read JavaScript? It varies, and behavior can change without notice. Google renders JavaScript, but you shouldn't assume the same of every crawler. Putting key content in the initial HTML avoids the question. ### Should I block AI crawlers? It depends on your goals. You can often allow search-related crawlers while disallowing training crawlers. Blocking search crawlers can remove you from those answers, so decide agent by agent. ## About Comppound Comppound runs SEO and AI search for B2B SaaS companies: a person sets the plan and reviews samples and every high-risk page, and AI speeds up production. - [Get the free report](https://comppound.com/ai-visibility-report): When buyers ask ChatGPT about your category, are you in the answer? - [See pricing](https://comppound.com/pricing): Want the content written, reviewed and published for you?