The Problem
Your customers have started asking ChatGPT and Perplexity for recommendations instead of Googling — and your competitors keep showing up in the answers while you don't.
A 12-person B2B SaaS company sells a project-management tool. Their traditional SEO is healthy: they rank on page one for several head terms and publish two well-researched guides a month. But when the founder types "best project management tool for agencies" into ChatGPT, Perplexity, and Google's AI Overview, three competitors are named and linked — and his company isn't mentioned once. Support has noticed prospects arriving with the line "I asked an AI and it suggested [competitor]." Nobody on the team can explain why the AI ignores them, because the usual SEO dashboards (rankings, backlinks, Core Web Vitals) all look fine. AI search is a different game with no Search Console to tell you what's wrong.
The Solution
Run a Generative Engine Optimization (GEO) audit instead of a traditional SEO audit. GEO scores the signals AI engines actually use to pick which sources to cite — whether AI crawlers can even reach your pages, how "liftable" your passages are as a direct answer, whether your site has an llms.txt, and how strong your entity signals (Wikipedia, schema, brand mentions) are.
Use the generative-engine-optimization skill to produce a 0–100 GEO Score with a six-category breakdown and a ranked fix list. Then use the schema-markup skill to implement the JSON-LD entity signals the audit flags, and the seo-audit skill to keep the traditional foundations (crawlability, HTTPS, sitemap) solid underneath. The biggest wins are usually fast and free: most sites are accidentally blocking AI crawlers and have no llms.txt.
Step-by-Step Walkthrough
1. Run the GEO audit
Tell your AI agent:
Run a GEO audit on https://acmepm.com. We want to get cited when people ask ChatGPT and Perplexity for "best project management tool for agencies." Give me a GEO Score and a prioritized fix list.
The generative-engine-optimization skill fetches the raw (pre-JavaScript) HTML, checks robots.txt, looks for llms.txt, inspects schema, and scores citability. You get:
GEO Score: 47 / 100 (Poor — AI systems may struggle to cite this site)
AI Citability & Visibility 38/100 GPTBot blocked; no llms.txt
Brand Authority Signals 55/100 Reddit + LinkedIn present; no Wikipedia
Content Quality & E-E-A-T 62/100 Strong guides, but no author attribution
Technical Foundations 71/100 SSR ok; sitemap missing from robots.txt
Structured Data 30/100 Organization schema only; no sameAs/Person
Platform Optimization 45/100 No question-headings or direct-answer blocks
2. Unblock AI crawlers (highest ROI, ~5 minutes)
The audit flags the single most damaging issue first:
Your
robots.txthasUser-agent: GPTBot→Disallow: /, inherited from a 2023 scraping-protection config. ChatGPT cannot read any of your pages.
Apply the fix it gives you:
# Allow AI search crawlers
User-agent: GPTBot
Disallow:
User-agent: OAI-SearchBot
Disallow:
User-agent: PerplexityBot
Disallow:
Sitemap: https://acmepm.com/sitemap.xml
3. Add an llms.txt
Generate an llms.txt for acmepm.com from our top pages.
The skill crawls the site and writes a root-level llms.txt:
# Acme PM
> Project management software for creative and marketing agencies — task boards, time tracking, and client billing in one tool.
## Docs
- [Pricing](https://acmepm.com/pricing): Plans, per-seat pricing, and the free tier.
- [Agency Guide](https://acmepm.com/guides/agency-pm): How agencies run client work in Acme PM.
- [vs Competitors](https://acmepm.com/compare): Feature comparison against the main alternatives.
## Optional
- [Blog](https://acmepm.com/blog): Articles on agency operations and project workflows.
4. Rewrite key passages to be citable
The audit shows the "What is agency project management?" intro is 90 words, pronoun-heavy, and buries the answer. Ask:
Rewrite that intro as a self-contained answer block following GEO citability rules.
You get a ~150-word passage that opens with a definition ("Agency project management is…"), answers in the first sentence, and includes two concrete stats with attribution — the exact shape AI engines lift into an answer.
5. Implement the entity signals
Hand the schema gaps to the schema-markup skill:
Add Organization schema with sameAs links to our LinkedIn, G2, Crunchbase, and YouTube, and Person schema for each guide author.
This raises the Structured Data sub-score and gives AI models the entity links they use to associate "Acme PM" with the agency-PM topic.
6. Re-audit and track the delta
After deploying, re-run the audit. The GEO Score moves from 47 to the mid-70s — crawlers unblocked, llms.txt live, citable passages and schema in place. Re-run it monthly: AI search platforms change their sourcing behavior, and new competitor content shifts the landscape.
Real-World Example
An agency-focused SaaS discovers via the audit that GPTBot has been blocked since 2023 and they have no llms.txt. They unblock the crawler, ship an llms.txt listing their pricing/comparison/guide pages, rewrite the intros of their five highest-traffic guides into self-contained answer blocks, and add Organization + Person schema with sameAs links. Their GEO Score goes from 47 to 78. Six weeks later, the founder's test prompt — "best project management tool for agencies" — returns an answer in both Perplexity and ChatGPT that now names and links their product, and support starts logging inbound prospects who say an AI assistant recommended them.
Related Skills
- generative-engine-optimization — The core GEO audit: citability scoring, AI crawler access, llms.txt, and platform optimization.
- schema-markup — Implement the Organization, Person, and Article JSON-LD the GEO audit recommends for entity recognition.
- seo-audit — Keep the traditional SEO foundations (crawlability, indexation, Core Web Vitals) solid beneath your GEO work.
- content-strategy — Plan the topic and authority coverage that earns AI citations over time.