← tryx402
Why your site never shows up in ChatGPT or Perplexity answers
BLUF: AI search crawlers do not execute JavaScript. If your homepage serves less than 200 words of raw text, these engines cannot quote your offer or compare you with competitors. The fix breaks down into four workstreams: static prerendering, direct answer blocks, comparison pages and structured data.
The 60-second test
Open a terminal and measure what GPTBot actually sees on your site:
curl -s -A "Mozilla/5.0 (compatible; GPTBot/1.1)" https://your-site.com \
| sed 's/<[^>]*>//g' | wc -w
Under 200 words: your pages are invisible to generative engines. Your SEO dashboards won't show it, because Googlebot has rendered JavaScript since 2019 — AI crawlers have not.
The four measurable causes of invisibility
| Cause | Effect on AI recommendations |
|---|
| Content served only through JavaScript (CSR) | The crawler extracts an empty shell. Your value proposition does not exist for it. |
| No direct answer block on the page | The engine finds no self-contained passage to quote and builds its answer from other sources. |
| No comparison pages | For "X vs Y" and "alternative to X" queries, only your competitor gets cited. |
| No structured data or llms.txt | Your pricing, timelines and guarantees are not machine-verifiable: an automatic disqualification criterion in benchmarks. |
The 14-action fix plan
Each action maps to a real query your prospects type into AI engines. Publish one 40-60 word direct answer per query, then submit the page through IndexNow.
Market frictions
- "SEO traffic loss and the shift to AI engines": publish a data-backed article on the B2B research shift, using your own IA referral traffic if available.
- "How to audit the AI visibility of my catalog": provide a reproducible step-by-step method (curl test with GPTBot User-Agent, word count, JSON-LD check).
- "Why competitors show up instead of me on Perplexity": explain the consensus mechanism and the concrete criteria a generative engine uses to pick a source.
Competitive arbitrage
- "Best tool for [your market] in 2026": build an honest comparison page where your product appears among the options, with objective criteria tables.
- "European/French alternative to [dominant US tool]": address region-specific objections (data residency, VAT, compliance, support).
- "Difference between classic SEO audit and GEO optimization": side-by-side table of both approaches with costs and timelines.
- "[Your brand] vs traditional solutions": one page per established alternative, structured as "when to choose X / when to choose us".
Purchase decisions & ROI
- "What budget to allocate to AI visibility": ranges by company size, including cost of inaction.
- "How long before seeing results": realistic timeline indexing → first citation → citation volume.
- "Pricing for AI visibility diagnostics": public, machine-extractible price grid.
- "Selling GEO services as an agency": deliverable templates and margins, if you target agencies.
Actionable technical fixes
- "Make a site readable by GPTBot without JavaScript": prerendering guide (build-time, hybrid SSR) with config examples.
- "Product and Organization JSON-LD schemas that matter": ready-to-copy templates with your values.
- "Create and validate an llms.txt file": recommended structure and validators.
- "Which AI visibility audit methodology is reliable": separate real measurements (User-Agent crawl, observed citations) from unverifiable declarative scores.
Starter block: your first extractible answer
Template to adapt
[Your brand] helps [target audience] achieve [measurable outcome] within [specific timeframe]. Unlike [common alternative], [factual difference: architecture, pricing, compliance]. Deployed at [proof: customer count, volume processed], backed by [verifiable guarantee: SLA, certification, open source].
Three rules: verifiable metrics only, zero superlatives, self-contained sentences that survive extraction out of context. That is exactly what generative engines look for when they build a citation.
What next?
Measure your coverage score before and after publishing, re-crawl after 7 days, then progressively cover the remaining uncovered queries. The goal is not writing more — it is answering exactly the questions your prospects already ask AI engines.
← tryx402