I'm using traditional SEO methods, but they don't seem effective for AI platforms: what to do differently
Why traditional SEO methods stop working on AI platforms and what to change: prompts over keywords, citations over rankings, crawler access, and how to measure it.
If your rankings are fine and AI engines still never name you, you have not done anything wrong. You have been measuring and optimizing for a system that works differently. Traditional SEO methods are not useless here, since the same web is the raw material, but several of their assumptions break. This guide covers what changes and what to do about each one.
What breaks when you carry SEO habits into AI search
Keywords become prompts. People do not type "crm software pricing" into ChatGPT. They write a paragraph about their team, budget and constraints. Engines also split that into several sub-searches, called query fan-outs, before composing an answer. Your page has to match a cluster of related questions, not one phrase.
Rankings become citations. There is no position seven. An answer names a few brands or cites a few sources, and everything else is invisible. The metric that replaces rank is whether you are named and cited, and in what position.
Clicks are no longer the only outcome. An engine can recommend you without sending a visit, or send a visit you never see attributed. Standard analytics often lump AI-referred traffic into direct or referral.
Your own site is not the only source. Engines lean on Reddit threads, YouTube videos, review sites and third-party lists. If you only optimize your pages, you are working on part of the evidence.
Crawler access becomes a first-order issue. AI bots such as GPTBot, OAI-SearchBot, ClaudeBot and PerplexityBot are not Googlebot. A firewall rule that blocks them is invisible in classic SEO reports.
What to do differently
1. Build a prompt set and track it
Write down 30 to 50 questions your buyers ask, grouped by topic and intent: category questions, competitor comparisons and brand questions. Track them on a schedule across the engines your buyers use. Promptwatch helps with the starting point: its Prompt Explorer suggests prompts from Search Console, ranking keywords, People Also Ask or any page, with search volume and difficulty on each.
2. Study fan-outs and cover the sub-questions
Look at what the engine searches for behind the scenes. Promptwatch shows query fan-outs per model, and its free GEO Inspector Chrome extension shows ChatGPT fan-outs and fetched versus cited sources. Then check that your page answers those sub-questions plainly, near the top, in text a crawler can read.
3. Fix crawler access first
Check robots.txt for the right bots, check firewall and CDN rules, and view page source to confirm content is not locked behind JavaScript. Promptwatch's Agent Analytics shows IP-verified AI crawler logs with a crawl-to-citation path per page, so you see whether the bots reached the page and what came back. Crawler logs start on the Professional plan.
4. Look at citations, not just mentions
When you are absent, check what the engine cited. Promptwatch's citation analytics break it down by page, domain, Reddit, YouTube and LinkedIn, and a Page Tracker lets you follow specific URLs, including competitors'. The action then follows the source: a better page, a spot on a third-party list, or a presence on a forum.
5. Refresh and publish against the gaps
Content Refresh in Promptwatch gives diff-style update suggestions for existing pages. Its Answer Gap analysis shows prompts where competitors are cited and you are not. On Business, the Content Agent plans, writes and publishes on a schedule with a review inbox.
6. Attribute the traffic
Visitor analytics attribute AI-referred visits and conversions through a lightweight script, a GTM template or the WordPress plugin, so you can see whether the work pays.
How the tools compare for this switch
| Tool | Starting price (as published) | Prompts and fan-outs | Citations | Crawler logs | AI traffic |
|---|---|---|---|---|---|
| Promptwatch | Free; paid from $95/mo | Volumes, difficulty, fan-outs | Page, Reddit, YouTube, LinkedIn | Professional and up | Yes |
| AthenaHQ | From $295/mo | Unlimited prompts, 11 models | Action Center guidance | Not documented | Not documented |
| Scrunch AI | From $250/mo | 125 prompts on Core | Sentiment, insights | AI bot traffic reporting | Not documented |
| Ahrefs Brand Radar | With paid Ahrefs; Index from $199/mo | Custom Prompts quota | Sources reported | AI crawler activity | No |
If you already use Ahrefs, Brand Radar's Custom Prompts is a gentle way to start, with a daily quota of 5, 10 or 20 prompts depending on plan. If you want to go from measurement to fixes and traffic in one product, Promptwatch is the better fit, because the crawler, citation and conversion layers are not documented in the others.
What stays the same
Good SEO fundamentals still pay: crawlable pages, clear structure, accurate schema, real expertise and honest content. AI engines read the same web. You are adding a layer, not replacing the foundation.
A realistic expectation
No change guarantees an engine will start naming you. Answers vary by model, country and phrasing, and some prompts will not move however good your page is. The practical aim is to measure where you stand, remove technical blockers, close the gaps that evidence points to, and check again next week.