GEO vs SEO: what actually changes for a small business

GEO stands for generative engine optimisation. It is the work of making your pages easy for AI assistants such as ChatGPT, Perplexity, Gemini and Google’s AI Overviews to find, understand and quote. You will also see it called AEO or AI SEO. The market around it is loud and the evidence is thin, so this post keeps three lists apart: what is documented, what is only reported, and what is guesswork.

Is GEO different from SEO?

For Google, barely. Its documentation on AI features in Search says that there are no additional requirements to appear in AI Overviews or AI Mode, and no other special optimisations are necessary. A page must be indexed and eligible to be shown in Google Search with a snippet. That is all it asks.

For the other assistants, each has its own crawler and its own published rules. They do not share one rulebook.

What GEO adds in practice is attention to three things that ordinary SEO often skips:

1. A clear identity. Can a machine tell who you are, what you do and where, from your own pages?
2. Text that is easy to quote. Short answers, one claim per sentence, honest dates.
3. Tracking. No report from Google or OpenAI tells you that you were cited, so you have to look yourself.

What is documented?

These are statements the companies make about their own systems. They describe eligibility. They do not tell you what gets chosen.

Claim	Source	Status
Google’s AI Overviews and AI Mode need only an indexed page that is eligible for a snippet	Google, AI features	Documented
No special schema.org markup is needed; structured data should match the visible text	Same page	Documented
No new machine-readable or AI text files are needed to appear in these Google features	Same page	Documented
nosnippet, data-nosnippet, max-snippet and noindex limit what Google shows from a page	Same page	Documented
OAI-SearchBot is for ChatGPT search, GPTBot is for training, and each setting is independent. OpenAI recommends allowing OAI-SearchBot	OpenAI, crawler documentation	Documented
Google-Extended does not affect inclusion in Google Search and is not a ranking signal	Google, common crawlers	Documented

What is reported but not proven?

• ChatGPT search draws partly on Bing’s index. This is widely reported. We did not find it in OpenAI’s own documentation. Being indexed in Bing costs nothing, so we do it anyway.
• Mentions on other sites matter. Directories, forums and review sites seem to shape how assistants describe a business. SEO practitioners report this, but none of the AI companies documents it.
• Answer-first, short, dated text gets quoted more. It fits how assistants quote pages. We have not seen proof that it causes a citation.
• Original numbers get cited. Practitioners say so. It is plausible, and it is a reason to publish only numbers you measured.

Treat these as reasonable bets. They are cheap, and they are also good writing.

What is guesswork?

• llms.txt. Google’s documentation says you do not need new machine-readable files or AI text files for its features. No major AI company has confirmed that it reads llms.txt. We publish one because it costs almost nothing, and we do not count on it.
• Hidden “AI-friendly” text, prompt stuffing, or content served differently to bots. There is no basis for it, and it can break search policies. We would not risk it.
• Special schema “for AI”. Nobody has shown that it exists.
• Any promise of a citation. Answers change from one run to the next. A guaranteed “#1 in ChatGPT” is not something anyone can sell honestly.

What does our own checklist look like on aaautomations.com?

We checked our own build on 31 October 2026. Everything below can be seen in the page source or at the URL shown.

Item	What the site does	Where to look
Text in the HTML	Every page is prerendered. Our previous site was a React app that served crawlers a 5 KB empty shell. The home page HTML is now about 80 KB	View source on any page
One URL per language	Separate /el/ and /en/ pages, with hreflang for el, en and x-default, in the page head and in the sitemap	/sitemap.xml
A clear identity	Organization data with the legal name, G.E.MI. number, VAT number, founding date and address	Home page JSON-LD
Page types	Service and BreadcrumbList on service pages, SoftwareApplication for our products, Article on work pages	JSON-LD on each page
Answer first	Each page opens with one or two sentences that answer its main question; headings are questions where natural	Any service page
Visible FAQs	Real questions with plain answers, shown on the page	Any service page
Crawlers	robots.txt allows all crawlers, AI crawlers included, and lists the sitemap	/robots.txt
llms.txt	A short company summary and page list	/llms.txt

Two notes on that table. We show FAQs because readers use them, but we do not mark them up as FAQ data: according to Google’s documentation, the FAQ rich result is no longer shown in Search. And allowing training crawlers is a choice. We allow them because we want to be known.

What we have not done yet:

• The sameAs list in our Organization data is empty, because we have no LinkedIn page or directory profiles yet.
• Our three products’ own sites do not yet say that we build them.
• We have no data on whether any assistant cites us. We have no client case study with citation results either.

The first two are the subject of our post on the name problem.

What should a small business do first?

In this order, because each step is cheap and builds on the one before:

1. Get indexed in Google and Bing. Verify the site in Search Console and Bing Webmaster Tools, and submit the sitemap.
2. Check that nothing blocks you. Look at robots.txt, your CDN or firewall rules, and any “block AI bots” switch your host turned on for you.
3. Put the facts on one page. Name, what you do, where, for whom, how to contact you, and prices or how you quote. Use the same wording in your profiles.
4. Open each page with the answer. Then the detail.
5. Keep dates honest. Change “updated” only when the content changed.
6. Be listed where customers look. Real profiles only. No bought reviews.
7. Ask the assistants. Write down the ten questions your customers ask, put them to each assistant once a month, and record the result.

How would you know it worked?

You build the measurement yourself.

• A fixed panel of questions, asked monthly in each assistant with the same wording. Log the date, the assistant, whether you were mentioned and which URL was cited.
• Referrals from chatgpt.com, perplexity.ai and similar in your analytics. Many AI visits arrive without a referrer, so this is a minimum.
• Crawler requests in your server or CDN logs. If a crawler never visits, fix access first.

Look at the trend over months. Do not judge a change after a week.

Where can we help?

If you want the basics checked on your site, see our AI search (GEO) service or our wider SEO work. We will not promise a citation. We can fix what is documented, test the rest and measure it. You can contact us with your site address.

In short

• For Google’s AI features, the documented requirement is the same as for search: be indexed and eligible for a snippet.
• GEO adds a clear identity, quotable text and your own tracking. It is not a separate secret.
• Reported but unproven: Bing’s role in ChatGPT search, and the value of mentions on other sites.
• Guesswork: llms.txt, hidden text, special “AI schema”, and any promised citation.
• We apply this checklist to our own site and say what is still missing.