AI-search visibility audit and 30-day plan

Scope note: You didn't give me a domain, and I haven't inspected your site. The audit below is a checklist with the likely failure modes for a four-locale B2B SaaS site. Each check needs a pass or fail from a crawl and a prompt test before you treat it as a finding. Nothing here guarantees rankings or citations. AI answers are non-deterministic, and being cited isn't the same as being recommended.

1. Audit framework

A. Access and rendering (blocks everything else)

B. Multilingual integrity

C. Extractability, by page type

D. Structured data and machine-readable files

• JSON-LD that matches the visible content: Organization, SoftwareApplication or Product with Offer, FAQPage only where real FAQs exist, TechArticle or Article on docs, and BreadcrumbList.
• Check inLanguage and the localized priceCurrency in each locale.
• llms.txt pointing to the key product, pricing and docs pages per locale. Optionally a /pricing.md. Google says neither is required for AI Overviews. Treat them as low-cost help for other agents.

E. Authority and third-party presence

• Positioning on G2, Capterra, LinkedIn and Crunchbase should match your About page. Models look for consensus.
• Original data, named authors and sourced statistics on the pages you most want cited.
• Local-market presence: German and Japanese industry publications, and Chinese-language channels that matter for your buyers.

F. Measurement baseline

Build a set of 20–30 revenue-relevant prompts, split across the four languages, covering:

• category definitions
• "[brand] pricing"
• "[brand] vs [competitor]"
• integration questions
• how-to questions drawn from your docs

Run each prompt 3–5 times per platform (Google AI Overviews, ChatGPT, Perplexity, Claude, Gemini). Record the citation rate with its sample size, such as "cited 2/5". Record which URL and which locale got cited, and whether you were only cited or actually recommended.

2. Prioritized 30-day plan

Week 1: Unblock and baseline (highest leverage)

1. Crawl and render test all four locales for product, pricing and docs. Fix server-rendering gaps first.
2. Review robots.txt and WAF/CDN rules per crawler purpose. Decide deliberately on discovery vs. training bots, and document the policy.
3. Fix hreflang, canonicals and sitemaps, and remove locale auto-redirects that hide pages from crawlers.
4. Run the baseline prompt test and save the results. Without this baseline you can't tell later whether anything worked.

Week 2: Pricing and product pages (most commercial value)

5. Rebuild pricing pages as plain HTML tables with units, limits, inclusions, localized currency and tax notes, and a last-updated date. Don't leave price points only in images or JS widgets.
6. Rewrite the first 40–60 words of each product page as a standalone definition, with sections that answer one question each. Don't chunk pages artificially. Write for people and use normal headings.
7. Add or fix JSON-LD (Organization, SoftwareApplication/Product and Offer) so it matches visible content in every locale.
8. Reconcile facts across locales into one source-of-truth sheet covering plans, limits, integrations and compliance claims.

Week 3: Documentation and citability

9. Restructure the top 20 docs pages (by traffic or by support-ticket topics): question-style headings, a direct answer first, version and date labels, and text code samples.
10. Add evidence to key product pages: sourced statistics, named customer outcomes with permission, and author or reviewer attribution.
11. Native-review the top pages in ZH, JA and DE for terminology and for how local buyers actually phrase queries. Don't ship translation-only fixes.
12. Publish llms.txt with links per locale to product, pricing and docs. Add /pricing.md if your pricing is simple enough to keep in sync.

Week 4: Presence, measurement and iteration

13. Align third-party profiles (G2, Capterra, LinkedIn, Crunchbase) with the site's positioning. Ask real customers for reviews. Don't buy or fabricate them.
14. Pursue 2–3 credible local mentions per priority market, such as analyst briefings, partner pages and industry publications.
15. Re-run the prompt set with the same method and compare citation rates and cited URLs with the baseline. Segment by locale.
16. Set a monthly cadence: quarterly freshness updates for pricing and top docs, a monthly prompt check, and AI-referral tracking in GA4.

3. Expectations

• Google AI Overviews and AI Mode are built on core Search systems. No special markup is required. Indexability, quality and E-E-A-T are what matter.
• Other engines reward extractable structure and authoritative, up-to-date owned pages. Citation shares shift between model updates, so treat results as trends, not guarantees.
• A 30-day window is enough to fix access, structure and consistency. Changes in citation rates may take longer, and some won't be attributable to a single change.

4. What I need to turn this into a real audit

• Your domain, or a list of priority URLs per locale.
• Your top 10–20 queries and your main competitors.
• Whether you have an existing robots.txt, schema and llms.txt, and which CDN or WAF you use.

With those I can run the crawl-based checks and fill in the pass/fail results.