Most people asking this question are trying to work out one thing: is a GEO agency real work, or a rebranded SEO upsell with a new acronym?
Fair question. “Generative engine optimization” became a pitch deck term quickly, and the label is used for very different services. This operator's guide describes seven areas a GEO agency may inspect, what evidence each area can produce, and where the limits sit.
A useful GEO engagement observes how a brand appears inside selected AI answers, inspects the public evidence those answers use, improves controllable sources, and repeats the sample. It cannot force an independent provider to name, cite, or recommend the brand.
The public Outrigger AI Visibility Index analysed 95,392 data points across 1,000 businesses and five AI models. Its April 2026 baseline found 65.9% of those businesses were invisible across all five models. This is industry research from a separate joint venture, not an Xpand client result or a forecast for any brand.
Disclosure:Outrigger is a separate joint venture Joel co-founded with Andrew and may support dated observations when configured for an engagement. PressForge remains in final beta and may organize digital PR workflow when that work is in scope. Neither product controls model output.
The 7 Workstreams a Real GEO Program Runs
GEO can involve several connected workstreams. The useful sequence depends on the baseline, the buyer questions, and the evidence already available. Here are seven areas a business or GEO consultant may inspect, without treating them as a fixed package.
01
Baseline Your AI Visibility
Before work is prioritised, record a dated sample of the answers buyers may see. The sample can note whether the brand appears, how it is described, which competitors appear, and which sources are cited.
What the work actually is
Agree the buyer questions and supported systems before collecting the sample. For each answer, record the date, wording, brand context, and cited sources. Outrigger, a separate joint venture involving Joel House and Andrew, may support this observation when configured for an engagement. The output is a sample, not a permanent rank or an explanation of hidden model logic.
Possible scoped output
A dated answer sample and source record, if measurement is included in the agreed scope.
How you'd verify it
Keep the questions, systems, dates, and observed answers so a later sample can be compared under documented conditions.
Joel's take
Ask to see the evidence behind any visibility claim. The AI visibility audit guide explains what a useful dated sample should preserve.
02
Build the Entity Foundation
Entity work reduces avoidable ambiguity in the public record. It helps buyers and retrieval systems distinguish the business, but it does not force a recommendation.
What the work actually is
We reconcile how your business is represented across public sources an engine may retrieve: structured data (Organization, Person, Product, and Service schema) on your own site; consistent name, address, and description across your directory and profile footprint; eligible public knowledge sources; and clean sameAs links tying genuine profiles together. The goal is to reduce ambiguity. No operator can force every engine to return the same facts.
Possible scoped output
Approved corrections to owned schema and eligible public profiles, where the review finds a material mismatch.
How you'd verify it
Compare the corrected public facts with a later dated answer sample, allowing for variation between systems and prompts.
Joel's take
Identity errors are often observable and correctable on owned pages and eligible profiles. Fix material contradictions first, then recheck how the business is described without promising that the correction will change a recommendation.
03
Inspect and Strengthen Public Corroboration
Some answer samples cite independent sources such as publications, directories, reviews, forums, or comparison pages. Those observations can reveal where a brand's important public claims are well corroborated or poorly represented.
What the work actually is
Compare the cited URLs in the baseline with the business's existing public evidence. Eligible inaccuracies may be corrected, and credible editorial coverage may be pursued when there is a substantiated story. Xpand may use PressForge, which remains in final beta, to organize research, pitching, and follow-up. Editors decide what is published, and publication does not guarantee an AI citation.
Possible scoped output
A prioritised corroboration plan, plus approved corrections or outreach activity when those actions are in scope.
How you'd verify it
Record what changed, then compare a later answer sample without promising a citation or a fixed movement date.
Joel's take
On-site clarity and independent corroboration answer different questions. A useful review checks both, then chooses the action supported by the observed source pattern. Our link building guide covers the mechanics of earning editorial links when that work is appropriate.
04
Restructure Content for Extraction
Clear, self-contained passages can make important facts easier for people and retrieval systems to interpret. That does not mean a system will quote, cite, or reward a particular content shape.
What the work actually is
Priority pages may benefit from direct answers, accurate dates, attributable claims, useful comparison structures, and explicit audience context. Original research or practical guidance can add public evidence when the business can support it. This is where answer engine optimization and classic on-page SEO overlap, with the same requirement that every material claim remain correct when read on its own.
Possible scoped output
Approved improvements to the priority pages or evidence assets identified in the scope.
How you'd verify it
Confirm the visible page and schema agree, then record whether later answer samples retrieve or cite the changed source.
Joel's take
The trap is treating GEO as a volume contest. Fewer accurate pages that answer real buyer questions are more useful than a larger set of unsupported or repetitive pages.
05
Manage Reviews and Corroborating Signals
Reviews and third-party profiles can corroborate public facts and may appear in a retrieval or citation set. Their relevance varies by category, question, system, and date.
What the work actually is
Review the accuracy and recency of eligible profiles, the substance of genuine customer feedback, relevant association listings, and the third-party pages already describing the business. The goal is a consistent and supportable public record, not manipulation of a hidden recommendation system.
Possible scoped output
A profile and review-process action list, with approved corrections where the scope allows them.
How you'd verify it
Check the public profiles directly and compare any later answer sample without assuming causation.
Joel's take
This is where GEO reconnects with ordinary business credibility. Independent corroboration must be earned, and no agency can promise that it will produce an AI recommendation.
06
Fix Crawlability for AI Engines
Crawlability is a technical eligibility question. A blocked or client-only page may not be available to a crawler, but access alone does not guarantee retrieval, citation, or recommendation.
What the work actually is
Review the robots rules for the user-agents the owner intends to allow, confirm important content is available in the returned HTML, and validate structured data against the visible page. An optional crawler-specific file should be published only when it can remain synchronized and truthful. Xpand removed its own public LLM files because that maintenance standard was not met.
Possible scoped output
Technical findings and approved fixes for material access, rendering, or structured-data defects.
How you'd verify it
Fetch the page with the relevant user-agent and inspect the returned content, directives, and server logs where access is available.
Joel's take
Technical access is worth checking early because it is directly observable. Keep the conclusion narrow: eligibility can be tested, while source selection remains the provider's decision.
07
Monitor How Answers Change Over Time
AI answers can vary as systems, prompts, sources, and dates change. Repeated sampling can show direction, persistence, and new inaccuracies without creating a universal rank.
What the work actually is
If ongoing measurement is in scope, repeat the agreed question set under documented conditions and record changes in brand context, cited sources, competitor appearances, and attributable referral traffic. The cadence should be agreed for the engagement rather than presented as a universal standard.
Possible scoped output
A dated comparison and next-priority note when recurring measurement is included in the scope.
How you'd verify it
Confirm the comparison preserves the original questions, systems, dates, and raw observations.
Joel's take
Ask upfront which observations will be retained, how often the sample will be repeated, and what limits apply. Rankings and traffic can remain useful alongside answer-level observations, but they measure different parts of the buyer journey.
“An agency selling GEO without its own way to measure a model's answers is guessing. The measurement isn't part of the work: it is the work.”
Joel House · Founder, Xpand Digital
GEO Consultant vs GEO Agency: Which Do You Need?
“GEO consultant” and “GEO agency” are used inconsistently. Compare the named owner, scope, evidence standard, and implementation responsibility rather than relying on the label.
A GEO consultant commonly focuses on assessment, priorities, and decision support, with the client team owning implementation.
A GEO agency may also implement agreed content, technical, measurement, or digital PR work. The actual responsibilities should be written into the engagement.
Read the seven areas above and identify what the internal team can own. Then ask any provider to separate advisory work, implementation, third-party costs, approvals, and measurement. Xpand scopes the work around the identified constraint rather than treating every area as a required package.
How to Spot a GEO Agency That's Faking It
Because GEO is new and demand is running ahead of competence, the market filled with providers selling the acronym without the work behind it. These are the tells. Any one of them should make you ask harder questions; two or more and you're looking at a rebranded retainer.
No baseline measurement
The proposal makes visibility claims without preserving the questions, systems, dates, answers, and cited sources behind them.
“We check it manually in ChatGPT”
Manual review can be useful when its limits are stated. It becomes misleading when a small convenience sample is presented as a universal score.
A tactic chosen before the evidence
The proposal prescribes content, schema, or digital PR before inspecting the current answer and source patterns.
Rankings and traffic as the only report
Rankings and traffic remain useful SEO measures, but they do not directly record whether a brand appeared or how it was described in an AI answer.
Guaranteed placement in AI answers
No outside provider controls what an independent AI system outputs. A provider can commit to agreed work and evidence standards, not placement.
Tooling presented as proof
Named software is not evidence by itself. Ask what it records, which systems it supports, what its sample misses, and whether raw observations are available.
The through-line is inspectability. A provider should distinguish observation from inference, controllable work from provider-controlled output, and public evidence from a forecast.
The Questions to Ask Before You Hire
You don't need to be technical to separate the operators from the storytellers. Ask these five, in this order, and listen for specifics rather than reassurance:
Which buyer questions and systems will you observe? The answer should define the sample and its limits.
What raw evidence will you retain? Look for dates, answer text, brand context, and cited sources.
Which actions are controllable? Ask the provider to separate owned changes, editorial outreach, and independent model output.
What will you report, and how often? The cadence and comparison method should be explicit rather than implied by a package name.
What does this cost, and what is excluded? Ask for a written scope, ownership, third-party costs, and terms. The GEO agency page explains Xpand's current scope and limits.
What a GEO Agency Actually Costs
Pricing depends on the measurement scope, number of markets and systems, implementation ownership, content requirements, and whether digital PR is included.
An honest proposal states the work, evidence standard, exclusions, ownership, third-party costs, and terms. The GEO agency page explains how Xpand currently scopes the work.
The useful decision is whether AI answers materially affect the buyer journey and whether the baseline identifies a gap worth addressing. The public research establishes a broad visibility problem, not the value or likely outcome of a specific engagement.
The Wider Playbook
GEO is one part of how a business gets found in 2026. These are the pieces that connect to it:
GEO agency for Xpand's current scope and measurement boundaries.
A GEO agency can record how a brand appears across a dated set of AI answers, inspect the public sources those answers cite, improve controllable owned content and entity clarity, pursue credible corroboration when appropriate, and repeat the sample. The scope varies by business, and no agency can guarantee a citation or recommendation.
The disciplines overlap on crawlability, useful content, entity clarity, and credible authority. SEO commonly measures rankings, traffic, and conversions. GEO adds direct observation of AI answers, brand context, and cited sources across a dated question set.
A consultant commonly focuses on assessment, priorities, and decision support. An agency may also implement agreed content, technical, measurement, or digital PR work. Titles vary, so the useful comparison is the named owner, scope, evidence standard, and implementation responsibility.
Keep the agreed questions, systems, dates, answer text, brand context, and cited sources. Repeat the sample under documented conditions and compare directional movement, persistent gaps, new inaccuracies, and attributable referral traffic where analytics supports it. This is an observation, not a universal AI rank.
There is no universal timeline. Answers vary by prompt, system, source availability, and date. A useful engagement defines controllable implementation milestones and a recheck cadence, without promising when an independent provider will retrieve, cite, or recommend the brand.
Ask which buyer questions and systems will be observed, what raw evidence is retained, which changes are controllable, how public claims are substantiated, and how later samples will be compared. Reject any promise of guaranteed placement or access to hidden model weights.
Cost depends on the measurement scope, number of markets and systems, implementation ownership, content requirements, and whether digital PR is included. Ask for a proposal that states the work, evidence standard, exclusions, and terms rather than relying on an unsourced industry range.
Not necessarily. Existing SEO work may already cover technical access, content quality, entity clarity, and authority. Add a GEO-specific layer only when AI answers matter to the buyer journey and the team needs direct observation of brand context and cited sources.
A useful GEO program records what the buyer sees, improves controllable public evidence, and repeats the observation without claiming control of the model.
Joel House · Xpand Digital
Want to see where you stand inside the AI answers?
Tell us which buyer questions and competitors matter. We'll confirm whether a measured baseline is the right first move and how it should be scoped.
