What determines GEO scope: item by item.

The scope of a GEO project, meaning the work of putting in place the preconditions for a site to be cited as a source in answer engines, is determined not by a ready-made package but by the site’s current technical and content condition. There are eight items that make the scope larger or smaller: the technical foundation, the number of templates, the number of languages, entity architecture and external profiles, the citable structure of the content, the measurement setup, who implements the fixes, and the continuity of the work.

There are no amounts, timelines or guarantees on this page, and that is deliberate. No provider can guarantee that an answer engine will show you as a source, or in what position you will be mentioned for a question; answer engines choose their sources with their own systems, and that choice changes over time. What can be promised is that the preconditions are put in place and the result is measured honestly. The items below spell out, one by one, what makes the scope of those preconditions larger or smaller, so that you can read the proposals in front of you in the same terms.

In GEO consulting work, we set the scope in a written proposal based on the site’s current state; we do not publish a list of ready-made packages. The questions below are the very questions we ask when preparing that proposal. We explained the same logic for the website itself, item by item, in the guide what determines a website’s price.

Does GEO require a special file or markup?

For Google, no. Google’s documentation on its AI features states clearly that there are no additional requirements or special optimizations needed to appear in AI Overviews or AI Mode, and that no new machine-readable files, AI text files or markup are required. For a page to be shown as a supporting link in these features, it needs to be indexed and eligible to be shown in Google Search with a snippet; in other words, the basic conditions of classic search.

This is the most important adjustment to scope. Special files or tags offered under the name “GEO package” do not, on their own, change anything on Google’s side. The same goes for llms.txt: some AI agents read the file, some never request it; Google has not announced any such support. That is why llms.txt is a small item in the scope and not a guarantee; why and how it is written is covered in the guide llms.txt and robots.txt. The weight of the work lies in crawlable HTML, structured data, entity consistency and the structure of the content.

How does the site’s technical foundation change the scope?

Because most answer engines choose their sources from search indexes, GEO does not replace technical SEO; it sits on top of it. If a page cannot be crawled, carries noindex, has a canonical address pointing to another page, or its content only appears after JavaScript runs, work on citability has no foundation to rest on. What robots.txt says to AI crawlers and how the server actually responds to them are part of the same foundation.

• Increases scope → accumulated technical SEO debt: crawl blocks, pages that are not indexed, scattered canonical addresses, content generated only in the browser with JavaScript.
• Reduces scope → a sound technical foundation; then only the GEO layer is set up.

If the technical foundation is not sound, saying so in the first step of the audit and recommending fixing it first is a more honest route than inflating the GEO scope. Schema and passage structure built before the foundation debt is paid sit on top of a page that is not being read.

Number of pages or number of templates?

Most of the machine-readable layer is set up at the template level. If hundreds of product pages share the same layout, the structured data is designed once, coded once and tested once; pages that were each built by hand in a different layout, on the other hand, are handled one by one. That is why, for schema and the technical layer, what counts is not the number of pages but the number of distinct templates.

On the content side, the math works differently. The citability assessment is done page by page, but not for every page on the site: only for the pages that are candidates to answer the questions customers actually ask. What determines the scope is the number of those candidate pages and their current state. Even on a large site the list of candidate pages can be short; on a small site, the list of questions whose answers have never been written can be long.

Why is the number of languages a separate item?

Each language needs its own set of questions, its own entity information and its own structured data. When the same question is asked in different languages, answer engines may choose different sources; that is why measurement is also set up separately for each language. Linking the language versions to each other correctly, keeping the names and descriptions in the schema consistent in every language, and which language the external profiles speak are jobs that do not exist on a single-language site. As the number of languages grows, what grows is not the text but the burden of keeping everything consistent.

What state is the entity architecture in today?

Whether answer engines recognize a brand as a single entity rather than as scattered pages depends on what the site says about itself being consistent: organization name, address, founder, services, and how these are linked to external profiles. A site with no schema at all and a site with messy schema are two different jobs; in the second, the contradictions are first found and cleared out, then the gaps are filled. The concept itself is covered in the entity architecture guide.

External profiles are an often-skipped part of the scope. If the information in business profiles, social accounts and industry directories contradicts what is on the site, the fix is made outside the site and depends on access to those accounts. If the accounts are in the company’s hands, the work is shorter; if no one knows who holds them, the question of ownership is resolved first.

Is the content written in a citable way?

Answer engines take not a whole page but a piece of it. That piece has to stay correct even when it is pulled out of its context: a first sentence that is understandable on its own, headings written the way readers actually ask, and an answer that is not hidden behind an expandable box. The details of this structure are in the citable passage guide; on this site’s guide pages, too, the first sentence of the introduction being correct on its own is a rule checked in pre-publication testing.

The question that changes the scope most is often this: will existing content be restructured, will new content be written, and who will write it? Turning existing text into a question-and-answer layout and preparing pages from scratch for questions whose answers have never been written are different workloads. Will the company’s experts write the content, or will the provider prepare it? If this line is not written into the proposal, the scope is unclear too.

How will measurement be set up?

A GEO project that is not measured is a statement of faith. Measurement is built from four items, and each has its own scope:

• Fixed question set — questions customers actually ask that stay the same in every measurement; how many questions and how many languages there are determines the scope.
• Engine-by-engine observation — asking the same questions separately in engines such as ChatGPT, Google AI Overviews and Perplexity and recording the date, the sources cited and the context; the method is in the measuring brand mentions guide.
• Server logs — direct evidence of whether AI crawlers actually visit the site; access to the logs depends on the hosting and on any services in between, and how to read them is in the AI crawler logs guide.
• Structured data coverage — which template carries which schema and whether it passes validation.

Search Console is not enough on its own here: Google shows the traffic of pages that appear in AI features not on a separate line in the performance report but within the “Web” search type. That is why engine-by-engine observation and server logs are separate items; how often they are repeated is also part of the scope. A line that cannot be measured is written in the report as “could not be measured”; it is not filled in with an estimated metric.

Who will implement the fixes?

A GEO audit can end with a list of findings, or it can also cover implementing those findings. The difference is large: schema, entity information and passage structure often require changes at the template and source-code level. If the person who does the audit can also make the fix in the source code, the team does not change between finding and solution; if there is no access, the plan is handed over to the site’s developer and implementation is tied to their schedule. Both routes are legitimate, but which one is chosen should be written into the proposal. The general form of this distinction is in the comparison SEO agency or hands-on technical specialist.

The platform is also a variable here. If there are limits to touching the content management system’s template or plugins, some fixes can be made in the existing system and some cannot. The audit needs to separate these point by point: which is possible in today’s system, and which is waiting for a platform change.

A one-off job or an ongoing one?

GEO has a setup part and a monitoring part. The setup, meaning repairing the technical foundation, building the schema into the templates, making entity information consistent and restructuring the candidate pages, is done once. Monitoring, however, is ongoing: answer engines change how they choose sources, new pages are added to the site, schema needs maintenance. A single observation is not enough to reach a verdict; the pattern emerges from asking the same questions again at regular intervals.

Continuity is a separate line in the scope and often the least discussed. If it is not written down who will repeat the measurement after setup, how often, and who will fix any deviations found, the value of the initial effort wears away over time.

How do you compare two GEO proposals?

Often two proposals are not pricing the same job. One contains an audit report and an llms.txt file; the other, repair of the technical foundation, template-level schema, entity architecture, restructuring of content, engine-by-engine measurement and implementation. Both are called “GEO”. The safest way to compare them is to write the following lines side by side for the two proposals:

• The state of the technical foundation and the SEO debt to be paid first
• Number of templates and the templates where structured data will be set up
• Number of languages and each language’s own question set
• Entity architecture and external profiles
• Content to be restructured or written, and who writes it
• Measurement: question set, engines, server logs, frequency
• Who does the implementation, and access to the source code
• Continuity of measurement and maintenance after setup

A line left blank shows where the difference comes from. If a proposal guarantees rankings or citations, that shows something too: that a result that cannot be guaranteed is being sold.

In GEO there is no citation you can buy; there are preconditions you can put in place and a trace you can measure.

If you want to know how your brand is cited in ChatGPT, Google AI Overviews and Perplexity, our GEO consulting records which questions and engines cite you and builds the prerequisites.