GEO vs SEO: What Changes, What Stays the Same, How to Measure
Generative engine optimisation builds on SEO rather than replacing it. Here is what stays the same, what changes, where off-site presence fits and how to measure it.

AEO & GEO
To get cited in Google AI Overviews, your page has to be indexed, eligible to show a normal snippet, and contain a short, self-contained passage that answers one of the sub-questions behind the query. There is no special tag or AI-only markup to add. The work is answer-first formatting, schema that confirms who you are, coverage of the entities around the topic, and a simple citation log so you know whether any of it is working.
This is the playbook I use for the AEO and GEO programme across the brands I look after at Garage Collective. It runs in the order I would actually do the work, with a check at the end of each step. For the strategic picture, read GEO vs SEO; this post is the hands-on part.
AI Overviews are built on Google's normal search index. Googlebot crawls your page, Google indexes it, and when a query triggers an AI Overview, Google pulls passages from indexed pages to support the generated answer. The links shown with the answer point to those supporting pages.
nosnippet directive or a very low max-snippet value limits what can be shown from your page. Blocking Google-Extended in robots.txt does not remove you from AI Overviews; that token governs use of your content for Gemini, not Search.In my experience the cited pages usually already rank on page one for the query or a close variant, though not always in first position. So this sits on top of ranking: it decides which of the ranking pages gets quoted.
Before rewriting anything, rule out the boring problems. I find at least one of these on most sites I audit.
nosnippet, max-snippet:0 or tight max-snippet value in the meta robots tag or the X-Robots-Tag header.data-nosnippet attribute wrapping the content you want quoted. Some themes add it to whole sections.How to check it worked: re-inspect the URL, request indexing if you changed anything, and confirm the answer text appears in the crawled HTML. For a full-site sweep, use my technical SEO audit checklist.
In my experience this is where most of the gains come from. Google lifts passages, so every section should make sense if it is the only thing someone reads.
Write the H2 or H3 the way people phrase the question, then answer it in the first two to four sentences underneath. Detail, caveats and examples come after. If the answer is a number, a definition or a yes or no, say it in the first sentence.
Repeat the subject instead of leaning on "it" or "this". A passage that starts "It usually takes a week" means nothing out of context. Here is the kind of rewrite I make, using a made-up brand called Acme Books:
| Weak passage | Citable passage |
|---|---|
| It works with most of them. | Acme Books imports ledgers from Tally and syncs payments from Razorpay. |
| This depends on your plan. | Acme Books includes three user seats on the Starter plan and unlimited seats on Pro. |
Numbers with units, product names, prices in rupees, dimensions, named standards and ingredient percentages are easy to quote because they support a precise statement. General advice gets summarised; specific detail gets cited.
How to check it worked: copy each H2 section into a blank document on its own. If a stranger could not tell which product, place or topic it refers to, rewrite the first sentence.
Free checklist
39 checks to get your brand cited in AI Overviews, ChatGPT, Perplexity and Gemini
Google has said there is no special schema needed for AI Overviews, and schema alone will not get you cited. What it does is remove ambiguity: who wrote this, which organisation publishes it, which product a price belongs to, when the page last changed.
| Page type | Schema | Fields I always fill |
|---|---|---|
| Home or about page | Organization | name, url, logo, sameAs links to LinkedIn, Crunchbase, marketplace and social profiles |
| Every page | BreadcrumbList | The real hierarchy, matching the visible breadcrumb |
| Blog post or guide | Article or BlogPosting, with a Person author | headline, author with jobTitle and sameAs, datePublished, dateModified |
| Product page | Product with Offer | name, brand, sku, price, priceCurrency, availability; aggregateRating only if the reviews are on the page |
| Local or service page | LocalBusiness or Service | name, address, telephone, areaServed |
| Q&A section | FAQPage | Only questions and answers that are visible on the page |
Two rules I do not bend. First, the markup must match what a visitor can see. Marking up reviews or FAQs that are not on the page breaks Google's structured data guidelines and can lead to a manual action. Second, keep dateModified honest. Do not change the date unless the content changed.
FAQ rich results have been limited to a small set of well-known government and health sites since 2023, so FAQPage markup will not earn most sites the expandable result. I still use it where the questions are real, because it labels question and answer pairs clearly.
How to check it worked: run the URL through the Rich Results Test and the Schema Markup Validator, then check the relevant Search Console reports a week or two later for errors on the templates you changed.
An AI Overview usually answers more than the literal query. For "best accounting software for small business in India", the answer may touch on GST compliance, e-invoicing, pricing, Tally migration and mobile access in a few lines. Each of those is a sub-question, and each can cite a different source. Google has described this approach as "query fan-out" for AI Mode, and AI Overviews behave in a similar way in my experience.
Your page should cover the sub-questions and name the entities (products, standards, places, people, brands) that a complete answer would include. This is how I build the list for a target query:
This is the same keyword, intent and content-gap thinking I use for regular SEO, run at passage level. For MoneyGence, a B2B accounting software client, that meant comparison, alternatives, use-case and integration pages that each answer one job cleanly, instead of one long page trying to answer everything.
How to check it worked: go back to the claims you copied from the current AI Overview. For each one, can you point to a sentence on your page that supports it, or adds something the cited source does not?
Treat Search Console as your record of impressions and clicks, and keep a separate citation log. What you need to know is which queries cite you, which URL they cite, and who gets cited when you do not.
| Column | What to record |
|---|---|
| Query | The exact search, grouped by target page |
| Date and location | AI Overviews vary by location and over time, so note both |
| AI Overview shown | Yes or no |
| Your URL cited | Which URL, or no |
| Other sources cited | The domains, so you know who you are up against |
| Claim supported | The statement your page is backing up |
| Change made | What you edited on the page, and when |
I check a set of 20 to 50 priority queries every two weeks, by hand for the most important ones. Most of the big rank trackers now flag when an AI Overview appears and whether your domain is in it, which saves time on the long tail. Manual spot checks still matter, because a tool's check is one snapshot from one place.
In Search Console, watch the pages you changed. When impressions for a query climb while click-through rate falls, there is often an AI Overview answering it on the results page. Your log then tells you whether you are the source or someone else is.
Give each change four to six weeks before judging it, and on your priority pages change one thing at a time. Otherwise you will not know what worked.
If you use AI to help draft the rewrites, keep someone with real subject knowledge in the loop to add the specifics. I explain how I run that in using AI for SEO content without losing E-E-A-T.
No. Google says no special markup is required, and pages without schema get cited regularly. Schema helps by making authorship, organisation and product facts unambiguous, so treat it as a support layer rather than the main lever.
Free AI search audit
Send your site and our founder reviews it personally. You get a clear, prioritised read on what is holding your search growth back.
Keep reading
Talk to the founder
Tell us what you want to rank for and where growth has stalled. We will tell you straight whether search is the right lever, and what it is worth.
Abhishek DwivediFounder and Lead Strategist, AD Search Studio
