--- title: "How We Work: The Underneath Method and First 90 Days | Underneath" description: "How Underneath works: diagnose, architect, deploy and compound; what we look beneath the answer, the first 90 days, the weekly review and reporting." canonical: "https://underneath.agency/approach" published: 2026-09-26 updated: 2026-09-27 publisher: "Underneath (https://underneath.agency/agent)" entity: "https://underneath.agency/.well-known/entity.json" --- Approach · The Underneath Method # Understand first. Then say it *everywhere*. Publishing more pages does not make a business more visible in AI answers. When an AI system recommends a company, describes a product or cites a source, there is an information system underneath that answer: the entities involved, the claims made about them, the sources and evidence behind those claims, and what the system can retrieve. We work backwards from the answer to understand that system, then decide what needs to change. **Diagnose → Architect → Deploy → Compound.** That is the Underneath Method. 1. 01 · Weeks 1 to 3 ### Diagnose A baseline of how search and AI answers describe you, and why. 2. 02 · Weeks 2 to 4 ### Architect A prioritized plan, with owners and measures. 3. 03 · Month 2 onward ### Deploy The changes ship, on your site and beyond it. 4. 04 · Ongoing ### Compound Measure, learn and set the next cycle. Diagnose · Weeks 1 to 3 ## Find out what the market, and the machines, actually *understand* about you. We start with your customers, not with a list of pages to write. We examine how your business appears across search and AI surfaces, the questions buyers ask, the competitors they compare you with, and the information available about your company across the web. We look beyond mentions and rankings. We investigate the **claims, sources, entities and relationships underneath the answer.** ### We examine AI visibility : Where you appear, where competitors appear, how descriptions change across questions, and which answers and citations matter. Search visibility : How your existing search presence supports or contradicts the information AI systems encounter. Entity understanding : How your company, products, services, people and locations are represented and connected. Source ecosystem : Which first-party and third-party sources contribute to the story around your business. Customer language : What customers say in reviews, sales conversations, questions and comparisons, and how that differs from the language your business uses. Competitive information : What competitors are consistently associated with, supported by, cited by and understood for. Technical foundation : Whether the information you want discovered can actually be found, interpreted and measured. The result is a baseline and a set of priorities, not a generic SEO audit or a list of content ideas: a clear view of **what is understood, what is missing, what conflicts and what matters next.** Architect · Weeks 2 to 4 ## Decide what needs to become *true*. Diagnosis tells us where the gaps are. Architecture determines what we do about them. We turn the findings into a prioritized plan across the website, content, technical foundation, entities, sources and measurement. Every recommendation comes with the reason for it. ### We identify - What should be clearer - What needs stronger evidence - What information is missing - Which relationships need to be established - Which sources deserve attention - Which pages matter most - Which technical issues are blocking discovery - Which opportunities are worth pursuing now - What should wait The output is a practical roadmap with goals, priorities, owners and measures. **No activity for activity’s sake.** Deploy · Month 2 onward ## Change the information system around the *business*. The work follows the diagnosis, so it looks different for every company. ### We may work across Technical : The systems that allow important information to be discovered, understood and measured. Content : Pages and information designed around the questions, decisions and claims that matter. Entities : The way your company, products, services, people and locations are represented and connected. Sources : The wider information environment that supports the understanding of your business. Evidence : The material that makes important claims clearer, more credible and more referenceable. Measurement : The signals that tell us whether the changes are actually affecting visibility and business performance. We don’t prescribe the work before we understand the problem. Compound · Ongoing ## Every cycle makes the next one *smarter*. AI answers shift, search results move, competitors publish and your own business keeps changing. A one-time optimization project can’t keep up with that. So we measure what changed, investigate why, test our assumptions and use what we learn to decide what comes next: **Measure → Learn → Change → Test → Compound.** Each cycle starts from better evidence than the last. Over time, what builds up is not just more mentions but a stronger, more accurate information footprint around the business. The first 90 days ## From signature to shipped work, without a long *silence*. The clock starts when the program does. You see the audit in the first week, a plan within the month, and the first real work live in the second. 1. Day 0 ### Agreement and access Goals, markets, competitors, measurement and access are agreed. 2. Days 1 to 7 ### Audit delivered A baseline across search, AI visibility, the website, sources, entities, customer language and measurement. 3. By day 30 ### Plan agreed Priorities, pages, sources, technical work, owners and measures. 4. Days 31 to 60 ### First work live The first changes ship, with the baseline kept so they can be measured. 5. Day 90 ### First full review What changed, what we learned and what the next quarter should focus on. At no point does the work vanish into a strategy deck. Beneath the answer ## The answer is only the *surface*. An AI answer can look simple. The information behind it isn’t. We investigate these layers together. - ### The question What are buyers actually asking? - ### The entity Who or what is being discussed? - ### The claim What is being said about it? - ### The source Where does that information come from? - ### The evidence What supports the claim? - ### The relationship What other entities, products, people, categories or concepts are connected to it? - ### The answer How does all of that become the description, comparison, recommendation or citation a customer sees? In practice, we take the answers to your most important customer questions and work backwards: which companies appear and how they are described, which sources are referenced, what information shows up consistently, what is missing, where competitors have stronger evidence and where your own information disagrees with itself. **The answer tells us where to look.** The deeper analysis tells us what to change. Engagement model ## One team. *One* accountable lead. The team is shaped around the work: one plan, one accountable lead, and specialists brought in where the work needs them. Some specialist work is done by vetted partners, under the same lead and the same standards. - ### Strategy lead Owns priorities, decisions and the overall program. - ### GEO specialist Owns AI visibility research, analysis and ongoing experimentation. - ### Content specialist Turns the opportunities the diagnosis finds into clear, useful pages that other sources can reference. - ### Technical GEO specialist Handles the technical and information architecture that makes that information discoverable. Nothing in a black box ## The weekly 30. Every program has a 30-minute review in the same slot every week, with a decision-maker from your side in the room. Four things, always in this order. 1. 01 ### What shipped The work completed, with links. 2. 02 ### What changed The evidence across AI visibility, search and business metrics. 3. 03 ### What we learned What the data and research are telling us. 4. 04 ### What happens next The next priorities and decisions required. A written note follows within one business day. You should never have to wait for a monthly report to find out what your agency has been doing. If a decision-maker cannot join the weekly review, we may not be the right fit. Reporting ## A rhythm you can plan *around*. Weekly : The weekly 30 and its written note. Monthly : Performance against agreed goals, including visibility, citations, traffic, leads and other relevant business measures. Quarterly : A deeper strategy review and a revised plan based on what the previous quarter taught us. Always on : Shared visibility into the metrics and technical signals that matter. How we think ## Seven questions, asked in *order*. - ### Understanding Why do your best customers choose you? - ### Recognition Do search engines, AI systems and the sources they read recognize who you are? - ### Retrieval Can the right information be found when it matters? - ### Evidence Are the important claims supported? - ### Competition Where do competitors have a stronger information footprint? - ### Measurement What actually changed? - ### Learning What do we know now that we didn’t know before? The questions are simple. Answering them well is the work. Free strategy call ## Nothing is optimized in *isolation*. Your website is one source. Your reviews are another. Your customers, publishers, directories, communities, search results and other sources all contribute to the information environment around your business. Our job is to understand how those pieces fit together, then strengthen the ones that are letting you down. Bring the problem to a free 30-minute call and we’ll tell you which pieces we would look at first. [Talk through my problem](https://underneath.agency/contact)[See what’s included](https://underneath.agency/services/generative-engine-optimization) --- This is the Markdown twin of https://underneath.agency/approach. The HTML page is canonical. Publisher: Underneath, https://underneath.agency/agent. Site index: https://underneath.agency/llms.txt.