How do I do generative engine optimization (GEO) for my website?
A practical GEO workflow for your website: crawl access, indexing, prompt research, content structure, citations and measurement, with the tools that cover each step.
Generative engine optimization (GEO) is the work of making your website get named and cited in answers from ChatGPT, Google AI Overviews, Perplexity, Gemini and Copilot. The honest starting point is that most of it builds on ordinary SEO. Google's own guidance on AI features in Search points site owners back to the same fundamentals: crawlable, indexable pages with useful content. Where GEO differs is in what you measure and which sources you need to be present on.
Here is a workflow that works for a typical site, in the order we would do it.
Step 1: let the right bots in
If an AI crawler cannot fetch a page, that page cannot be cited from a live fetch. Check your robots.txt, CDN rules and bot protection against the bots that matter: GPTBot, OAI-SearchBot and ChatGPT-User from OpenAI, ClaudeBot, Claude-User and Claude-SearchBot from Anthropic, PerplexityBot and Perplexity-User, and Google's own crawlers. OpenAI documents that OAI-SearchBot is the one used for search results, so blocking it is the most direct way to disappear from ChatGPT Search. Decide on training bots separately from search bots, since they are different decisions.
Do not guess. Look at your server or CDN logs. This is where most tools stop and where Promptwatch starts: its Agent Analytics shows real-time, IP-verified AI crawler logs, filters out spoofed user agents, reports errors with failure explanations and shows a crawl-to-citation path with a citation rate per page. Crawler logs begin on the Professional plan at $245 a month.
Step 2: make sure pages are indexable and snippet-eligible
Standard hygiene still applies: pages return 200, are not blocked by noindex, load without hiding the main text behind scripts that crawlers skip, and have clear titles and headings. Check Google Search Console and Bing Webmaster Tools for indexing problems. Bing has also been previewing AI-related reporting in Webmaster Tools, so check its current documentation for what is available to you.
Step 3: research the prompts, not just keywords
People ask AI engines long questions, and engines break them into several sub-queries (fan-outs) before answering. Build a prompt list from your customers' real questions, grouped by topic and intent: what is, best for, versus, how much, how do I. Promptwatch's Prompt Explorer suggests prompts from Search Console, ranking keywords, People Also Ask or any page, and shows search volumes, difficulty scores and per-model query fan-outs.
Step 4: write pages that answer the question directly
Lead with the answer, then support it. Use specific facts, named entities, dates and original data where you have it. Keep one topic per page, use descriptive headings, and add structured data where it honestly matches the page. Do not add schema to chase a ranking factor Google has said it does not need for AI features. Content gap analysis helps here: in Promptwatch it scores your coverage against the prompts you track and shows which questions you have no page for.
Step 5: earn mentions off your site
Engines often cite third-party pages: Reddit threads, YouTube videos, review sites, industry publications. Check which domains show up in answers to your prompts, then work out where you can legitimately appear. Promptwatch has Reddit, YouTube and LinkedIn citation reports and an offsite mentions view for pages that name your brand without a link.
Step 6: measure, then iterate
Track visibility score, share of voice, sentiment and position in answer over time, and compare periods before and after a change. Tie it to traffic: Promptwatch's visitor analytics counts AI-referred visits and conversions from ChatGPT, Perplexity, Gemini, Claude, Copilot and others. Its Action Board turns the data into a prioritized to-do list.
Tools that cover each step
| Step | Promptwatch | AthenaHQ | Scrunch AI |
|---|---|---|---|
| Crawler logs | Yes, from Professional | Not published | Bot traffic reporting |
| Prompt tracking | All active models on paid plans | 11 models, from $295/mo | 4 LLMs on Core, $250/mo |
| Citation analytics | Page, domain, Reddit, YouTube, offsite | Not documented at that depth | Not documented at that depth |
| Content creation | Content Agent on Business | Content optimization agent | Optimized page serving |
| Traffic attribution | Visitor analytics with conversions | GA4 and Shopify connections | Not published |
AthenaHQ is a sound pick for a team that wants an action queue. Scrunch AI is interesting if the crawler-side serving approach appeals. Profound is a bigger enterprise program. If you only need content produced, tools like Outrank publish articles on a schedule, but they do not measure whether engines cite them, so you would still need a tracker.
Where to start this week
- Check robots.txt and CDN bot rules for the named bots above.
- List 30 prompts your buyers actually ask.
- Run a free baseline: Promptwatch's Explore tier covers 10 prompts on ChatGPT, and its AI Visibility Report and Agentic Readiness Check need no login.
- Fix the pages that matter most, then re-measure.
GEO is not a trick. It is the same careful publishing as before, plus a loop that tells you whether the engines noticed.