Ask ChatGPT "what's the best tool for X" and you get a shortlist of three or four names with a short justification each. For a growing number of buyers, that shortlist is the entire research phase.
If you are not on it, you never find out. There is no referrer, no impression count, no rank tracker line going down. You simply do not get considered, and your analytics show nothing at all. This is the blind spot that generative engine optimisation exists to address.
What are GEO and AEO, and how are they different from SEO?
Three overlapping jobs, often confused:
- SEO — ranking in classic search results. The user sees ten blue links and picks one.
- AEO (answer engine optimisation) — winning the featured snippet, the voice answer, the box at the top. The user sees one answer and often does not click.
- GEO (generative engine optimisation) — being named and cited inside an AI-generated answer from ChatGPT Search, Perplexity or Gemini. The user sees a synthesised recommendation.
They reward overlapping but not identical things. SEO rewards authority and links. AEO rewards a clear, self-contained answer placed near the question. GEO rewards being quotable, being consistently described the same way across the web, and being accessible to AI crawlers in the first place.
How do I measure whether AI mentions my brand?
You cannot fix what you have not measured, and manually asking ChatGPT the same question every week is not measurement — answers vary between runs, so a single sample tells you nothing.
The AI Brand Visibility Tracker queries ChatGPT, Perplexity and Gemini across the topics you care about and reports, per model:
- Whether your brand was mentioned at all
- Your ranking position within the answer — being named third matters less than being named first
- Sentiment — being mentioned as the cheap option is different from being mentioned as the good one
- Share of voice against the competitors named alongside you
- The sources the model cited — the single most actionable field on the list
That last one is where the strategy comes from. If Perplexity keeps citing three comparison articles on sites you have never contacted, you now know precisely where to spend your effort.
Why am I not being mentioned?
Usually one of four reasons, in rough order of how often they turn out to be the cause:
- AI crawlers cannot read your site. A robots.txt that blocks GPTBot, ClaudeBot, PerplexityBot or Google-Extended removes you from the training and retrieval pipeline. Plenty of sites block these by accident, having copied a config from a blog post.
- Your content is not quotable. Models lift self-contained paragraphs that directly answer a question. A page that buries the answer in paragraph nine, after a personal anecdote, does not get quoted.
- You have no structured data. FAQPage, HowTo and Product schema tell a machine what your page asserts rather than making it infer.
- Nobody else describes you. Models synthesise from many sources. If your own site is the only page that mentions your product, you look like a claim rather than a consensus.
The SEO, GEO & AEO Audit scores any site 0–100 per page across all three dimensions and specifically checks AI-crawler access, llms.txt presence and schema validity — which covers the first three causes in one run.
Audit your own domain first
Before optimising anything, find out whether AI crawlers can even reach you. One run, 0–100 scores per page. $5 free credit on a new Apify account.
Run the SEO/GEO/AEO audit → Create a free Apify account →What is llms.txt and do I need one?
llms.txt is a plain-text index of your site written for AI consumption, sitting at your root the way robots.txt does. Instead of an AI crawler parsing your navigation, cookie banner and footer to work out what you offer, it reads a clean list of your pages with descriptions.
It is a young convention rather than a settled standard, and no model publicly guarantees it uses one. But it costs one file, it cannot hurt, and the upside if adoption continues is meaningful. The llms.txt Generator crawls your sitemap and produces spec-compliant llms.txt and llms-full.txt files you upload to your root.
For a working example, this site publishes one.
What about local businesses?
If you serve a geographic area, AI answers increasingly lean on local search data, and local rank is not a single number — it varies street by street. A single rank check from your office postcode tells you almost nothing.
The Google Maps Geo-Grid Local Rank Tracker checks your local-pack position from many points on a grid around the business and reports per-point rank, average rank, coverage and Share of Local Voice. The typical finding is that you dominate within two kilometres and vanish beyond it — which is a targeting problem you can only see with grid data.
A practical order of operations
- Check crawler access. Open your robots.txt and confirm GPTBot, ClaudeBot, PerplexityBot, OAI-SearchBot and Google-Extended are not blocked. This is a five-minute fix with the largest possible payoff.
- Measure the baseline. Run the brand visibility tracker on your top ten buying-intent questions before you change anything, so you can tell later whether the work did anything.
- Fix the quotable-answer problem. On each key page, answer the page's implied question in the first paragraph, in complete sentences that survive being lifted out of context.
- Add FAQPage and HowTo schema where you genuinely have questions and steps. Do not fabricate an FAQ to get the markup — that is a well-known way to lose rich results entirely.
- Publish llms.txt. Cheap, quick, plausibly useful.
- Go after the cited sources. The tracker told you which pages the models quote. Getting mentioned on those is the highest-leverage work available.
- Re-measure monthly. AI answers drift as models update. This is a tracking discipline, not a one-off project.
Frequently asked questions
Indirectly, and slowly. You cannot edit the model. You can make your site readable to its crawlers, make your claims quotable, and increase how many independent sources describe you consistently. That is what shifts a synthesised answer over months — anyone promising to change it this week is selling something.
That is a real trade-off, not an obvious call. Blocking protects content from being reproduced; it also removes you from AI-mediated discovery entirely. If AI search is a plausible acquisition channel for you, blocking is choosing invisibility. If your content is the product, blocking may be correct.
Monthly is a sensible cadence for most businesses. Answers vary run to run, so any single check is noisy — you are looking for a trend across several months, not a reading on one day.
Honestly, unproven. No major model publicly commits to reading it. It is one small file, so the cost of publishing one is close to zero and the option value if adoption grows is real — but do not expect a measurable change from it alone.
No. Classic search still drives far more traffic for almost every site. GEO is a new surface to cover, not a replacement, and most of the underlying work — clear answers, real expertise, structured data, third-party mentions — improves both at once.
Find out where you stand
Two runs will tell you more than a month of speculation: audit your domain with the SEO/GEO/AEO auditor to check whether AI crawlers can reach you, then baseline your mentions with the AI Brand Visibility Tracker. Both live in the SEO and AI visibility category.
⚡ Run this without building it
These actors already do what this guide describes. Free $5 credit on a new Apify account.
Get the Free Web Scraping Toolkit
Join the newsletter and get my curated list of scraping tools, proxy comparison cheatsheet, and Python automation templates.