# How do agencies onboard a new GEO client step by step?

> Start GEO client onboarding with real buyer questions from sales and customer conversations, not a keyword export alone.

Source: https://bestgeoplatforms.com/geo-client-onboarding · Publisher: GEO Platforms for Agencies · Updated: 2026-11-13

## TL;DR
- Start GEO client onboarding with real buyer questions from sales and customer conversations, not a keyword export alone.[3]
- Freeze a dated AI visibility baseline before implementing fixes, and save the prompts, engines, markets, languages, answers, citations, and sources.[2]
- Agree the competitor set explicitly because vendor examples and plan limits are not industry-wide standards.[9][21]
- Audit crawl access, indexability, visible content, and crawler policies before proposing GEO-specific changes; Google does not require special AI schema.[18]
- Set a transparent reporting cadence and distinguish mentions and citations from clicks, leads, conversions, and revenue.

## How do you onboard a new GEO client?

Agencies should onboard a GEO client in six stages: define the business questions and scope, create a stable prompt set, capture a dated AI visibility baseline, agree the competitor set, audit technical access, and establish a reporting cadence. This order separates eligibility and measurement from later content work, while keeping citations, mentions, sources, traffic, and conversions as distinct outcomes.

A defensible **geo client onboarding** process does not begin with a generic keyword export or a promise of rankings. It begins with the questions buyers actually ask, records the exact engines and markets tested, preserves the original answers, and makes later changes comparable.[1][2]
## What should agencies collect during discovery?

Discovery should establish what the client wants to learn, which audiences and markets matter, which competitors should be compared, and which answer engines are in scope. AMEC guidance identifies prompts, audiences, markets, languages, competitors, platforms, tools, and dates as core measurement inputs.[1]

Use sales calls, customer interviews, support tickets, product research, and existing content briefs to identify real buyer questions. Jincove recommends using questions heard during sales calls because they represent questions that can precede purchase, rather than relying only on brand-heavy prompts.[3]

The discovery deliverable should be a signed scope sheet containing:

- Business objectives, such as increasing qualified discovery, improving product consideration, or monitoring reputation.
- Priority audiences, personas, markets, and languages.
- Products, services, locations, and entities to track.
- Answer engines and search surfaces to test.
- Direct and indirect competitors.
- Core sources or pages the client believes should support answers.
- Stakeholders responsible for SEO, content, web development, legal, communications, and reporting approval.
- Access requirements for analytics, Search Console, server logs, CMS, CDN, and relevant vendor platforms.
- Measurement limitations, including known regional, engine, prompt, or data-access constraints.

Do not describe an AI visibility baseline as historical if monitoring has only just started. AMEC states that baseline data begins when monitoring starts, so the initial measurement should be dated and treated as the starting point for prospective trend analysis.[4]

## How should an agency build the prompt set?

A useful prompt set combines branded and unbranded questions. Branded prompts test how an engine describes the client directly; unbranded prompts test whether the client appears when a buyer asks about a category, problem, use case, location, or comparison.

Keep the first version stable and versioned. Record the exact wording, engine, market, language, date, and any relevant settings for every prompt.[5][4]

### Numbered GEO client onboarding checklist

1. Collect real questions from sales calls, customer research, support conversations, and product teams.[3]
2. Group questions by topic, funnel stage, audience, product, market, and intent.
3. Separate branded, category, comparison, problem-led, local, and transactional questions.
4. Remove duplicate or unnatural prompts.
5. Mark which prompts are business-critical and which are exploratory.
6. Record the exact prompt wording rather than storing only a topic label.
7. Define the engines, locations, languages, and account settings used for each test.
8. Decide which prompts belong to the stable benchmark and which can change during research.
9. Add source and competitor fields so each answer can be reviewed later.
10. Version the prompt set before the first baseline run.

There is no universal prompt count established by the reviewed agency and measurement frameworks. The more important controls are balanced intent coverage, stable wording, documented scope, and transparent changes over time.[6]

Avoid making most prompts brand-specific. A brand-heavy set can show whether an engine recognizes the client, but it can miss the earlier category and problem questions that influence discovery.[7]

## How do you create an AI visibility baseline?

Run the agreed prompts before implementing fixes. Save the full answer, not only a screenshot or a single visibility score, and record the engine, market, language, date, competitor set, mentions, citations, and source URLs.

A practical **AI visibility baseline** should distinguish at least these measures:

| Measure | What to record |
|---|---|
| Presence | Whether the client appears in the answer |
| Mention | Whether the brand, product, or entity is named |
| Position | Where the client appears in the answer |
| Citation | Whether the client’s domain is cited |
| Sources | Which pages and domains are cited |
| Framing | How the client is described |
| Competitors | Which alternatives appear and how they are positioned |
| Accuracy | Whether claims are correct, incomplete, or misleading |
| Risk | Unsupported, outdated, or potentially harmful statements |
| Business outcomes | Clicks, leads, conversions, and revenue measured separately |

A citation is not the same event as a click, and visibility should not be reported as revenue without separate outcome evidence. Outcome tracking has its own gap: Finseo's [AI search attribution study](https://www.finseo.ai/blog/ai-search-attribution) found that analytics records only about 1 in 10 AI-driven conversions, and it covers how a post-purchase survey can close that gap. AMEC also recommends treating presence, framing, citations, omissions, source quality, accuracy, and risk as separate observations rather than collapsing them into one number.[1]

Repeat testing matters because AI answers can vary across runs, prompts, and time. A 2026 measurement paper recommends describing visibility as a distribution rather than relying on a single observation.[6] Preserve repeat outputs and disclose whether a result comes from one run, repeated runs, or aggregated monitoring.

Finseo can fit here as one platform option for agencies that want prompt tracking, competitor and source analysis, reporting, and downstream task workflows in the same system. For client access and data handling, its [enterprise page](https://finseo.ai/enterprise) lists SSO via SAML, roles per project, an audit trail, and unlimited seats. It also lists hosting in Frankfurt, Germany with EU data residency, GDPR compliance with a data processing agreement (DPA), and ISO 27001 certification. Its public product materials describe daily or weekly tracking, engine filters, competitor analysis, source views, reports, and agency-oriented features, but agencies should still document the exact prompts, engines, markets, dates, and platform settings used for each client.

## How should competitors be selected?

Competitor tracking should be agreed during discovery, not added casually after the first report. Ask the client to identify the competitors that buyers compare with them, then validate that list against category research and observed answer-engine results.

RankZero recommends defining direct competitors and using the client as a benchmark; its onboarding material gives three to four direct competitors as an example.[8] Semji documents a product limit of up to 10 competitors depending on plan.[9] Neither example establishes an industry-wide competitor quota.

For each competitor, record:

- Why the competitor is in scope.
- Whether it is direct, indirect, aspirational, local, or substitute competition.
- Which products or pages make the comparison relevant.
- Which markets and languages apply.
- Whether the competitor is tested in branded and unbranded prompts.
- Which sources appear to support its visibility.
- Whether the competitor is consistently present or appears only in isolated answers.

The baseline should show the client and competitors under the same prompt, engine, market, and date conditions. Otherwise, differences may reflect test design rather than actual visibility.

## What belongs in a GEO technical audit?

A GEO technical audit should first establish whether important content can be crawled, indexed, rendered, and understood. It should not promise that technical changes alone will produce citations. Rendering deserves particular attention: Finseo's own [audit of AI crawler readability](https://www.finseo.ai/blog/ai-crawler-readability) notes that most AI crawlers never run JavaScript and found that 8.9% of 20,293 audited pages on 524 sites returned nothing readable.

Google says that AI Overviews and AI Mode do not require special optimization, and that existing SEO fundamentals remain useful. Pages must be indexed and eligible to appear with a snippet in Google Search to be eligible as supporting links in these features.[10]

Check the following:

- Robots.txt access for relevant crawlers.
- Noindex, canonical, and redirect handling.
- Indexability of priority pages.
- Internal links to important content.
- Server responses, CDN rules, and authentication barriers.
- Rendering of visible text and key page sections.
- Structured data accuracy and consistency with visible content.
- Product, service, organization, author, and location information.
- Duplicate, thin, outdated, or contradictory content.
- Search Console coverage and enhancement issues.
- Whether important pages depend on interactions that crawlers cannot access.
- Whether source pages contain direct, extractable answers to buyer questions.

Google states that there is no special schema markup required for AI features and that structured data should match visible page content.[11] Treat claims about mandatory “AI schema” or a guaranteed GEO markup advantage cautiously.

### Crawler and policy checks

Record crawler policies separately because search inclusion, model training, and product controls are not interchangeable.

- Google Search controls crawling and indexing through established Search mechanisms; robots.txt controls crawler access, while `noindex` is used to prevent a URL from appearing in Search.[12]
- Google-Extended controls the use of crawled content for future Gemini training and grounding and does not affect Google Search or other Google products.[13]
- OpenAI separates OAI-SearchBot, used to surface websites in ChatGPT Search, from GPTBot, associated with training; the settings are independent.[14]
- OpenAI says a robots.txt change may take approximately 24 hours to be reflected.[14]
- `llms.txt` is a proposal, not a Google Search eligibility requirement.[15][16]

Ask the client’s policy owner to approve crawler settings. Do not change robots.txt, noindex directives, CDN rules, or training controls as part of onboarding without documenting the intended consequence.


GEO technical audit checklist for agencies

Confirm access to the client’s Search Console, analytics, CMS, CDN or server logs, and relevant reporting tools. Crawl priority pages, inspect rendered content, review indexability and canonicals, test internal discoverability, compare structured data with visible content, and document crawler-policy decisions separately for Google Search, Google-Extended, OAI-SearchBot, and GPTBot. Save the audit date, affected URLs, evidence, owner, recommendation, and approval status.



## How should reporting cadence be agreed?

There is no universal GEO reporting interval. Some platforms document daily monitoring, while others support configurable scheduled reports.[17][5] The agency should choose a cadence based on the client’s decision cycle, prompt volume, answer variability, and expected rate of change.

A reporting agreement should specify:

- Data collection frequency.
- Client report delivery date.
- Stable benchmark prompts.
- Rules for adding, removing, or editing prompts.
- Engines, markets, and languages included in each report.
- Whether repeat runs are used.
- Evidence retained for each observation.
- Definitions for mention, citation, position, source, and visibility.
- Separate treatment of traffic, leads, conversions, and revenue.
- Owners for technical, content, reputation, and reporting actions.

A weekly report can be useful for active programs, while a monthly executive report may suit clients with slower publishing and approval cycles. These are operating choices, not universal standards. The key requirement is to disclose the schedule and keep enough of the benchmark stable to support comparison.[5]

### What should the report contain?

A concise client report can use this order:

1. Executive summary of material changes.
2. Measurement scope and any prompt or engine changes.
3. Visibility, mention, citation, and source trends.
4. Competitor movements.
5. New or rising questions and query fan-outs where available.
6. Technical access and indexability findings.
7. Content and source recommendations.
8. Risks, uncertainties, and answer variability.
9. Actions, owners, deadlines, and next review date.
10. Business outcomes reported separately from visibility.

Do not present a citation as proof of traffic, trust, or revenue. AMEC recommends connecting measurement to awareness, trust, behavior, and impact while distinguishing those outcomes from visibility itself.[1]

## What should the GEO client onboarding timeline look like?

A two- to four-week onboarding is a practical operating model for many agencies, but the exact duration depends on access, stakeholder availability, market scope, and prompt volume. The table below is a recommended workflow, not a published industry standard.

| Timing | Agency activity | Client deliverable | Output |
|---|---|---|---|
| Days 1–2 | Kickoff and discovery | Business goals, markets, audiences, competitors, access owners | Signed scope sheet |
| Days 3–5 | Prompt research and review | Buyer questions and priority topics | Versioned prompt set |
| Days 6–7 | Measurement setup | Engine, language, location, and account confirmation | Test configuration |
| Days 8–10 | Baseline run | Approval to freeze the starting point | Dated baseline and saved answers |
| Days 11–13 | Competitor and source analysis | Competitor rationale and priority pages | Comparison and source map |
| Days 14–17 | Technical audit | CMS, CDN, Search Console, analytics, and crawler-policy access | GEO audit findings |
| Days 18–20 | Recommendation workshop | Owners, priorities, and implementation constraints | Prioritized action plan |
| Days 21–24 | Reporting setup | Cadence, recipients, definitions, and escalation route | Reporting agreement |
| Day 25 onward | Ongoing monitoring | Feedback on changes and business outcomes | Recurring reports and re-tests |

Freeze the baseline before making changes. If the prompt set, engine mix, location, language, or tracking method changes later, record the change in the report rather than presenting the new result as a clean before-and-after comparison.[17]

## Which onboarding mistakes should agencies avoid?

The most common errors are measurement and scope errors rather than formatting errors.

- Starting with a keyword export and ignoring real buyer questions.[3]
- Using mostly branded prompts and calling the result category visibility.[4]
- Reporting one answer as a durable trend despite answer variability.[5]
- Changing prompts without versioning them.
- Comparing competitors under different engines, markets, or dates.
- Treating crawler access as a guarantee of citations or rankings.[18]
- Selling special AI schema as a mandatory prerequisite.[18]
- Treating Google-Extended as a Google Search visibility control.[19][13]
- Treating OAI-SearchBot and GPTBot as the same policy decision.[14][5]
- Reporting mentions, citations, clicks, leads, and revenue as one metric.[2]
- Claiming a universal prompt count or reporting cadence without evidence.[5]
- Making content recommendations before checking whether priority pages are crawlable and indexable.[20][18]

The strongest onboarding system is explicit about what is measured, what is not measured, what changed, and what the evidence can support.

## FAQ
### What is GEO client onboarding?
GEO client onboarding is the process of defining a client’s AI-search scope, creating a prompt set, capturing a dated visibility baseline, selecting competitors, checking technical access, and agreeing how results will be reported. It creates the measurement and delivery foundation for later content and technical work.[14]

### How many prompts should a GEO audit include?
The reviewed frameworks do not establish one universal prompt count. Agencies should choose enough branded and unbranded questions to cover the client’s priority audiences, products, markets, topics, and intents, then document the denominator and keep a stable core for trend analysis.[4]

### What should an AI visibility baseline measure?
An AI visibility baseline should record presence, mentions, citations, positions, cited sources, competitor appearances, framing, accuracy, omissions, and risk. Clicks, leads, conversions, and revenue should be measured separately because a citation and a click are different events.[3]

### Do GEO clients need special AI schema or an llms.txt file?
Google says there is no special schema markup required for AI Overviews or AI Mode, and structured data should match visible content.[11] `llms.txt` is an optional proposal rather than a Google Search eligibility requirement.[15][16]

### Should agencies block GPTBot while allowing ChatGPT Search?
That depends on the client’s policy, but OpenAI documents OAI-SearchBot and GPTBot as separate controls. Agencies should confirm whether the client wants ChatGPT Search inclusion, training exclusion, both, or neither before changing robots.txt.[14]

### How often should GEO reports be sent?
There is no universal reporting cadence. Daily monitoring and configurable scheduled reports are documented platform approaches, but the agency should choose a schedule that matches the client’s decision cycle and disclose prompt, engine, market, language, and repeat-test changes.[5][17]

## Sources
[1] The AMEC GEO Principles — https://amecorg.com/wp-content/uploads/2026/05/AMEC-GEO-Principles.pdf
[2] AI visibility measurement methodology · Aive GEO — https://www.aive.com/en/resources/methodology
[3] The GEO Audit Checklist for Agencies in 2026 | Jincove — https://jincove.com/blog/geo-audit-checklist-for-agencies-2026
[4] amec conflict.inddc — https://amecorg.com/wp-content/uploads/2026/05/26-05-15-A-Practitioners-Guide-to-GEO-Measurement.pdf
[5] Why We Measure AI Visibility Weekly — and Repeat Every Prompt Three Times | GEO Tracker AI Blog — https://geotrackerai.com/blog/why-we-measure-ai-visibility-weekly
[6] [2604.07585] Don't Measure Once: Measuring Visibility in AI Search (GEO) — https://arxiv.org/abs/2604.07585
[7] GEO: Generative Engine Optimization — https://dl.acm.org/doi/epdf/10.1145/3637528.3671900
[8] Onboarding | RankZero Docs — https://www.rankzero.io/docs/onboarding
[9] GEO configuration: set up your AI visibility monitoring — https://help.semji.com/en/geo-configuration
[10] Search generative AI control - Search Console Help — https://support.google.com/webmasters/answer/16908024?hl=en
[11] AI Features and Your Website | Google Search Central  |  Documentation  |  Google for Developers — https://developers.google.com/search/docs/appearance/ai-features
[12] Robots.txt Introduction and Guide | Google Search Central — https://developers.google.com/search/docs/crawling-indexing/robots/intro
[13] Google's common crawlers | Google Crawling Infrastructure  |  Crawling infrastructure  |  Google for Developers — https://developers.google.com/search/docs/crawling-indexing/google-common-crawlers
[14] Overview of OpenAI Crawlers — https://developers.openai.com/api/docs/bots
[15] /llms.txt—a proposal to provide information to help LLMs use websites – Answer.AI — https://www.answer.ai/posts/2024-09-03-llmstxt.html
[16] The /llms.txt file, v2 – llms-txt — https://llmstxt.org/?swcfpc=1
[17] Scheduled Reports and Monitoring · Geolity — https://www.geolity.ai/docs/monitoring
[18] GEO for Agencies: The Playbook for a New Service Line — https://maveriko.com/blog/geo-for-agencies
[19] Google's common crawlers | Google Crawling Infrastructure | Crawling infrastructure | Google for Developers — https://developers.google.com/crawling/docs/crawlers-fetchers/google-common-crawlers
[20] GEO Monitoring — https://docs.gradial.com/docs/insights/geo-monitoring
[21] Prompts — https://docs.geosurge.ai/tracking/prompts
