Uncategorized

What to Put in a GEO Engagement Scope: Deliverables, Reporting and Red Flags

Most GEO proposals promise visibility in ChatGPT, Perplexity and Google’s AI answers. Few say what the provider will actually do each month or how success is measured. This article sets out what a sound scope of work for Generative Engine Optimization (GEO) should contain, using public evidence on what can and cannot be controlled.

Key Takeaways
  • Google’s documentation says there are no additional requirements or special optimizations to appear in AI Overviews or AI Mode, so any promise of guaranteed placement is suspect.
  • SparkToro’s study found under a 1 in 100 chance that ChatGPT or Google’s AI returns the same brand list twice, so position-based reporting is unreliable.
  • Scope deliverables you can control: content, technical access, measurement design and review points.
  • Report visibility as a percentage across a fixed prompt set, repeated over time, not as a rank.

What no provider can promise

Google’s Search Central page on AI features states there are “no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary.” To be eligible, a page must be indexed and eligible to appear in Search with a snippet, and that does not guarantee it will be shown. The page also says you do not need new machine-readable files, AI text files or special schema.org markup.

Outputs from chat assistants are even less stable. SparkToro, with Gumshoe.ai, published a study on January 28, 2026 in which 600 volunteers ran 12 prompts through ChatGPT, Claude and Google’s AI results, producing 2,961 responses in November and December 2025. It reported under a 1 in 100 chance that ChatGPT or Google’s AI would return the same list of brands in two responses.

The implication for your contract is direct. A provider can improve the conditions for being cited, but cannot guarantee a citation, a position or a fixed share of answers. A scope that promises one is describing luck.

Deliverables worth scoping

Because outcomes are uncertain, a good scope names the work the provider controls. Each deliverable below can be checked as done or not done, independent of what any AI engine does next month.

DeliverableWhat it coversHow you verify it
Baseline visibility auditA fixed prompt set tested before any changesRaw responses and run counts delivered to you
Crawler and indexing reviewrobots.txt rules, rendering and indexability of key pagesDocumented before-and-after checks
Content updatesPages that answer buyer questions directly, with sourced claimsPage list, publish dates and change log
Third-party presenceWhere your brand is discussed outside your own siteList of target sources and outreach actually sent
Measurement setupReferral tracking for AI sourcesWorking analytics segments you can open yourself

The technical items connect to our guides on which bots to allow in robots.txt and JavaScript rendering and AI crawlers. The third-party item reflects the evidence summarized in brand mentions versus backlinks.

How reporting should work

SparkToro’s authors argue that the percentage of responses in which a brand appears across many prompts is a reasonable metric, while any tool that reports a ranking position in AI answers is unreliable. In their data, one prompt’s top brands appeared in 55 to 77% of 994 responses in a headphone follow-up, even though list order almost never repeated.

The same study found that 142 volunteers writing their own headphone prompts produced very different wording, with an average semantic similarity score of 0.081. That means your prompt set must reflect how real buyers phrase questions, not only the provider’s preferred keywords.

  1. Agree the prompt set and its source in writing before work begins.
  2. Require each prompt to be run many times per reporting period, with the run count shown.
  3. Report visibility as a percentage of runs, split by engine.
  4. Track referrals and inquiries from AI sources in your own analytics, not only the provider’s dashboard.
  5. Review trends monthly and treat a single reading as noise.

Our guide to measuring AI search visibility when answers change every run goes deeper on the method.

Red flags in a proposal

A proposal that treats a moving target as a fixed one is the main warning sign. Watch for guaranteed citations or rankings in ChatGPT, which Google’s documentation and the SparkToro data both argue against. Be cautious of a “secret” file or markup that supposedly unlocks AI answers, since Google states none is required. Ask what evidence supports llms.txt or schema claims before paying for them.

Reports that show a single ranking position, or a score with no raw responses behind it, cannot be checked. A provider who will not share prompts, run counts or response logs is asking for trust you have no way to verify.

If the provider also proposes autonomous agents that edit your site or accounts, add limits to the scope. Our article on least-privilege access for AI marketing agents and the agentic vendor evaluation checklist cover what to require.

Frequently Asked Questions

Should a GEO contract guarantee citations in ChatGPT?

No. Outputs vary from one response to the next, and SparkToro found under a 1 in 100 chance of an identical brand list twice. Scope the work instead.

How long should a GEO engagement run before judging it?

Long enough to collect repeated runs across several reporting periods. A single month shows noise as much as signal, so agree the review points in advance.

Is GEO different from normal SEO work?

They overlap heavily. Google says standard SEO fundamentals still apply to AI features, and our comparison of GEO, AEO and SEO shows where they diverge.

A clear scope protects both sides, and it makes the first month of a project easier to judge. At Grow with Hafsa we help US and UK businesses define measurable GEO and AI search work, and we build agentic workflows with approval points, as part of our AI marketing and performance marketing services.