We've worked in search since it was ten blue links.
Search now happens across three surfaces at once: traditional organic results, answer engines including Google's AI Overviews, and assistants like ChatGPT, Perplexity, Gemini, and Claude. The same technical foundation still decides most of it, which is why we run all three as one program rather than three retainers. Some clients hire us for the whole thing, and some hire us for a single audit that tells them how they're being described by the assistants their buyers use.
Twenty-seven years of watching search change, and one foundation that hasn't.
We came out of the web standards movement, so semantic structure, accessible markup, and clean URLs were part of the practice long before clients were paying for either. That turns out to be most of what answer engines reward. The tools changed, the vocabulary changed twice, and the underlying work of making a page easy for a machine to read, trust, and quote has been the same job the whole time.
Optimizing for search engines since the web standards era, through every major shift in how people find things.
Two higher education organic programs still running, one of them our longest at seventeen years.
Engaged sessions from ChatGPT, Perplexity, and Gemini doubled on a long-running higher education program.
Founded
Web standards and accessibility
Machine learning in production
AI Overviews change the results page
Prompt research and share-of-voice tracking as standard practice
Four situations bring most people to this page.
Someone asked ChatGPT about your company and it had nothing useful to say.
This is how one of our current engagements started. The CEO of a national restoration franchisor typed his own company name into ChatGPT and Claude, got back almost nothing, and a website project that had been deferred for years became a priority that quarter. It's a fast, unpleasant, and reasonably accurate test, because if the assistants can't describe you, the people asking them about your category are hearing about someone else.
Impressions keep climbing and clicks don't.
A national nonprofit came to us drawing 5.8 million Google impressions a year and about 103,000 clicks, which is a click-through rate under two percent at an average position of 10.6. They already ranked for the queries that mattered and sat just low enough to miss most of the traffic. That gap is usually a content and structure problem rather than an authority problem, and it's often the cheapest win available.
Traffic is falling and the content is fine.
Organic sessions decline, nothing on the site got worse, and the answer is that the answer now arrives before the click. We see this most in categories built around questions, where a resource library that took years to write is being summarized by an assistant that doesn't cite it. The fix is structural, and it usually applies at the template level across hundreds of pages at once.
A hundred location pages that all say the same thing.
Multi-location and franchise networks tend to generate near-duplicate pages with a city name swapped in, which competes against itself, thins out crawl budget, and gives an assistant no reason to name any one location. Meanwhile the Google Business Profiles sit unclaimed or half-finished, which is where the map pack traffic actually comes from.
SEO and AEO aren't two disciplines, and the tactics aren't as different as the headlines suggest.
The fundamentals have held up better than the discourse around them. Clean crawlability, semantic structure, fast pages, authoritative content, and honest internal linking are what earn a ranking, and they're also what make a page retrievable and quotable by a model. We treat traditional SEO practice as inherently AEO-supportive, then layer the AI-specific work on top where it actually pays. Anyone selling you a separate AEO retainer is selling you the same technical foundation twice.
That said, four things genuinely are different, and a program that ignores them is behind.
Prompt research alongside keyword research
Our SEO lead works with your team to compile 25 to 50 representative prompts reflecting how your buyers actually ask, then tracks responses over a baseline period to establish where you're cited today and which competitors are capturing share of response. Keyword volume still tells us what to build; prompts tell us how the question gets phrased when there's no search box in front of it.
Content structured for extraction
Question-formatted headers with direct 40 to 60 word answers underneath them, definition blocks, comparison tables, and summary sections are what assistants lift and attribute. This is a formatting discipline applied to content you may already have, which is why it's usually the highest-return work in the first ninety days.
AI crawler access as a deliberate decision
Your robots.txt has to say something explicit about GPTBot, ClaudeBot, PerplexityBot, Google-Extended, Applebot-Extended, and CCBot, and most sites are currently saying it by accident. We'll walk you through what each one does and where an llms.txt file helps.
Measurement that counts citations, not only clicks
Share of voice in AI Overviews and assistant answers gets tracked as its own metric, because a citation that never produces a click still shapes a decision. Reporting that only counts sessions will tell you this work isn't paying off while it is.
SEO is what gets you ranked. GEO is what gets you quoted.
The work covers six areas, and you can engage us on one of them or all of them.
Most programs pull from several. Each phase is scoped and priced on its own, and a one-time audit with a prioritized roadmap is a perfectly good way to work with us.
01 Technical SEO and site health Crawlability, indexation, Core Web Vitals, redirects, canonicals, and rendering. +
Our team verifies that search engines and AI crawlers can reach and correctly interpret your most important pages, which means checking redirect chains, robots.txt directives, canonical integrity, status codes, JavaScript rendering, and the crawl budget issues that quietly suppress visibility. Page performance is a direct ranking factor and an indirect quality signal to the models, so Core Web Vitals get treated as a requirement with a target rather than a report card. Where the site is large enough to justify it, our team runs log-file analysis to see what the crawlers are actually doing instead of inferring it.
02 Content architecture and clean-up Inventory, consolidation, pruning, and converting content the models can't read. +
A large share of the opportunity on most sites is fixing what already exists. Authority gets spread across duplicate and competing pages, older content dilutes the site's focus, and thin pages absorb crawl attention that should be going somewhere else. Our audit produces a page-by-page inventory with a keep, merge, or remove decision for every URL, plus a before-and-after view so the impact is visible. High-value content locked in PDFs is largely invisible to assistants, so we flag those for conversion into structured HTML that can be parsed and cited.
03 Keyword and prompt research Head terms, question-shaped queries, SERP features, and the prompts your buyers actually type. +
Our team identifies the terms worth winning and maps them to specific pages so nothing on the site competes with itself, covering head terms, long tail, and the question-based queries that surface in AI answers. Alongside that, we analyze SERP features (featured snippets, People Also Ask, and AI Overviews) to see what the results page for each term actually looks like, since a term with a strong AI Overview needs a different page than a term with ten links. Where the business case needs it, our team builds a bottom-up traffic projection by cluster and ranking position rather than a percentage pulled from the air.
04 Structured data and answer formatting Schema by page type, implemented in templates, plus the content structure the models quote. +
Schema is the clearest signal available for telling a machine what a page is and what it means. Our team implements and expands the types that fit your content (FAQPage, Organization, Person, Article, Product or SoftwareApplication, HowTo, BreadcrumbList, and LocalBusiness for location pages), inside your code templates rather than bolted on with a plugin. On a resource library or a program catalog, this is one templated change applied across hundreds of entries at once, which makes it the single most productive change on most sites we inherit.
05 Authority, E-E-A-T, and digital PR Author attribution, expert bylines, trust signals, and earned coverage. +
Google's Experience, Expertise, Authoritativeness, and Trustworthiness framework is increasingly mirrored by how models decide which sources to quote. Our team implements author bios, reviewed-by attribution, and executive summaries on technical content so both readers and models can see who stands behind it, and we tighten internal linking from your strongest hub pages so authority concentrates where you want it. On the off-page side we manage ethical, editorially-earned coverage. What we won't do is commit to a fixed number of links per month, because a quota produces the kind of links that used to work and now create risk.
06 Local, multi-location, and franchise search Google Business Profiles, citations, map pack visibility, and location pages that differ. +
For organizations with a geographic footprint, the Google Business Profile is usually the single highest-return asset and usually the least maintained. Our team audits and builds out categories, service descriptions, service areas, photos, Q&A, and review response practices, then pushes consistent name, address, and phone data across the major directories so Google can confirm you're a real local business. On the site side, location pages need genuine local differentiation (real team information, coverage areas, reviews, photos, and locally relevant detail) rather than a template with variables swapped in, and we'll build the content model that makes that sustainable for a distributed network.
Measurement, because a program you can't check is a program you have to take on faith.
Our analytics team owns GA4 and GTM implementation, server-side migrations, and offline conversion import so closed-won outcomes flow back into the picture rather than stopping at the form fill. On this service specifically, that means tracking AI referral traffic as its own channel, monitoring share of voice across your target prompts, and reporting rankings with analysis attached rather than a spreadsheet of positions. For higher education, we've been joining Slate inquiry records to GA4 behavior since 2013, so admissions teams can see which pages and channels produced applicants.
Beyond the Cookie: The 2026 Executive Playbook for Privacy-First MeasurementOur SEO lead is in the room during discovery, not after launch.
Most of what determines organic performance gets settled while a site is being designed and modeled, long before anyone writes a page. That work is covered in the next band, because it applies whether you hire us for search or hire us to build the site.
If you're hiring us to design or build the site, the search work happens during the build.
Plenty of clients come to us for a redesign or a replatform and have no intention of buying an SEO retainer. The search work still has to happen, because a site's organic performance is mostly determined by decisions that get made in discovery and content modeling, and the cost of revisiting them climbs steeply once templates exist. So our SEO lead is staffed onto design and development engagements from the start with a defined allocation of hours in discovery and again in implementation, and the recommendations land as build requirements with acceptance criteria rather than as a document handed over at launch.
Discovery The URL audit and the redirect map Built at the start of the project, not in launch week. +
Our SEO lead runs a complete inventory of your existing URLs at the start of the project, maps every one to its destination in the new site, and flags the pages carrying rankings, links, and conversions that the new architecture has to protect. The redirect map is a discovery deliverable, not a launch-week task, which is the single biggest predictor of whether organic traffic survives a replatform. This is also the phase where we settle the structural questions that are expensive to reverse later, including subdomain versus subfolder for any secondary sites, and how a multi-site or multi-location network links together so authority flows across it rather than pooling in one tier.
Information architecture Structure is a search decision before it's a design decision Navigation depth, URL taxonomy, heading hierarchy, internal linking. +
Navigation depth, URL taxonomy, heading hierarchy, and internal linking all get decided while the sitemap is being drawn, and all four are search decisions. Our content strategist and SEO lead work the same sitemap, which usually means flattening a deep navigation that has been burying your best material and consolidating pages that currently compete with each other. Where the research shows demand for a topic you don't have a page for, that page enters the sitemap during IA rather than arriving as a content request eighteen months after launch.
Content modeling The fields decide what schema is possible The part most teams discover too late. +
Structured data can only describe content the CMS has actually separated out. FAQ content stored as one rich-text blob can't produce FAQPage schema, while the same content stored as paired question and answer fields can, and the same logic applies to authors, dates, prices, locations, credentials, and program details. Our team models the content so entries carry the fields an answer engine needs in order to quote them, and we make the title tag, meta description, canonical, and schema fields required in the model, so an author can't publish a page missing them. Retrofitting any of this later means touching every entry on the site.
Design Templates that hold up when they're full Reviewed by designers and the SEO lead together. +
Our designers and our SEO lead review templates together for the things that only show up at scale: whether the H1 is the page's actual subject, whether body copy can carry the depth a competitive page needs, whether the FAQ and comparison modules exist as components, and whether a location or program page has room for genuine local detail rather than a variable swapped into boilerplate. Accessibility work overlaps heavily here, since semantic heading structure, meaningful link text, and alt text serve screen readers and crawlers with the same markup.
Development Rendering, schema, and performance as acceptance criteria Written into the same tickets as the components. +
Two implementation details decide more than anything else on this list. The first is rendering: content that only appears after client-side JavaScript can quietly fall out of the index, which is a live risk on headless and API-fed builds, so our engineers confirm that priority content renders server-side or statically. The second is schema, implemented inside the code templates (Article, FAQPage, Organization, Person, Product, BreadcrumbList, and LocalBusiness where locations are involved) rather than bolted on with a plugin, so it applies across every entry at once and survives the next design change. Core Web Vitals targets, crawl depth, sitemap generation, and internal linking patterns are written as acceptance criteria in the same tickets as the components themselves.
Migration and launch Tested against the real URL list Not a sample, and indexation staged in waves where it needs to be. +
The 301 map gets deployed and tested against the full live URL inventory rather than a sample, and where the new site is much larger than the old one (a franchise network moving from a handful of pages to a hundred location pages, for instance) we stage indexation in waves with quality criteria gating each release, so the site doesn't ship a few hundred thin near-duplicates on day one. After launch our team watches crawl behavior, indexation, and position for a defined window instead of assuming it held. A dip of a few weeks is normal. A permanent loss is a redirect failure, and it's avoidable.
Handover Your team gets the rules, in writing URL conventions, schema by template, internal linking logic. +
Editor training covers the search-relevant fields and why they're required, and the written documentation includes the URL conventions, the schema each template emits, and the internal linking logic. That's what keeps the structure intact after we're gone, which matters more on a distributed content operation than any single optimization.
Here's how an engagement actually runs.
We won't recommend content before the audit is done. That sounds obvious, and it's the most common place these programs go sideways, because content written on top of a broken structure inherits the problem.
The early work is diagnostic and corrective, which is why our projections show little movement in the first quarter and gains building across the rest of the year. Any agency showing you a straight line from month one is selling you a chart. We'll tell you which recommendations pay off inside a quarter and which take six to twelve months, because those are different buckets and pretending otherwise sets up a conversation nobody enjoys in month four.
What this looks like when it's working.
Three different tools, all surfacing the same school for the prompts that decide a shortlist. Drawn rather than captured, since we can show the results but not the institution's name.
total search impressions, at 11.5M year over year.
average search position, improved from 31.1, which is a move from page four to the top of page two.
AI-referred engaged sessions, from 1,203 to 2,408, meaning visits from ChatGPT, Perplexity, and Gemini doubled.
non-branded click-through rate, from 0.54% to 0.63%, which is traffic from people who weren't already searching the brand name.
A private university in the Northeast, February through April 2026 against the same period a year earlier. Organic visibility and AI visibility moved together, which is the evidence that the structured content work is doing both jobs at once. Source: Google Search Console and GA4.
Visits from ChatGPT, Perplexity, and Gemini, doubled year over year.
Higher is better. The line rises because the position number falls.
lift in organic traffic for a global life sciences manufacturer following a modular CMS rebuild, alongside a 140 percent increase in mobile conversions.
Search behaves differently by industry, and we've spent a long time in a handful of them.
The technical foundation is the same everywhere. What changes is who's searching, what they're deciding, and which surface the decision actually gets made on.
Higher education
Our longest-running organic engagements are in higher ed, including Adelphi University since 2013 and St. Joseph's University New York since 2017, alongside website and search work for Sarah Lawrence College, Northeastern University, and Gettysburg College. Prospective students now shortlist schools inside AI tools before they ever reach a .edu domain, which matters more here than in most categories because between 50 and 80 percent of applicants research and eliminate schools anonymously before they raise a hand (source: EAB), and roughly 78 percent of education-related Google searches return an AI Overview (source: EducationDynamics 2026 Modern Learner Report). We also close the loop on the enrollment side, joining Slate inquiry records to GA4 behavior so admissions teams can see which pages, campaigns, and keywords produced applicants rather than sessions.
Higher education →Non-profit
Our nonprofit search work covers mission, program, and donation content, where the visibility problem is usually that a large and genuinely authoritative content library is being summarized by assistants without attribution. We've worked with the Central Park Conservancy, Results for America, the Posse Foundation, Vitamin Angels, and the American Astronomical Society, and our current organic programs include a national Jewish philanthropic organization and a youth mental health nonprofit whose resource center reaches more than 250,000 people a month. Donation and campaign pages get protected explicitly during any content clean-up, because they convert quietly and they're the easiest thing to break.
Non-profit →B2B and professional services
Long sales cycles make organic search compound differently here, since the same page has to serve a first-time researcher and a buying committee doing diligence eighteen months later. We rebuilt West Monroe's site around their actual positioning and then took over their search engine marketing, and a global life sciences manufacturer saw organic traffic rise 32 percent after a modular CMS rebuild. Our work also runs deep in professional services, where the service and practice-area page structure is most of what decides organic performance.
E-commerce and consumer brands
Category, collection, and product page architecture decides most e-commerce organic performance, and the AI shopping surfaces are moving faster here than anywhere else. Our clients include Juliska, Marlo Beauty Supply, 39DollarGlasses, Fireclay Tile, and Morphe, and the work spans product schema, faceted navigation and canonical handling, branded search tracking, and Google Business Profile and Merchant Center reputation. We've also handled the harder end of that, including rebuilding review equity for a retailer after a Google Business Profile suspension.
E-commerce →Publishing and media
Publishers feel the shift to answer engines first and hardest, because their content is exactly what the models summarize. We've done platform and search work for the Observer, Commercial Observer, Playbill, Canary Media, McGraw Hill, Thomson Reuters, and the Associated Press. The questions that matter here are which crawlers to allow, how to structure content so citations point back to you, and how to keep an archive of thousands of articles discoverable rather than crawl-starved.
Publishing and media →Multi-location, franchise, and local
For franchise networks and multi-location businesses, the map pack and the Google Business Profile usually drive more revenue than the website does, and the location pages have to differ from each other in ways a template alone won't produce. We're currently building a three-tier network for a national restoration franchisor covering roughly 100 franchisees across territories in the US, Canada, and the UK, with location page templates, GBP build-out per location, and a governance layer that keeps corporate marketing in control of what publishes. Our restaurant and QSR work (&pizza, CAVA, Taco Mac, Colectivo Coffee) runs on the same local fundamentals.
We implement what we recommend, which is rarer in this discipline than it should be.
Most SEO engagements end at the document. You get an audit, a spreadsheet of recommendations, and a quarterly call, and then the recommendations sit in a queue behind everything else your development team is doing. A year later the audit gets run again and finds the same issues. We're a full-service agency with 87 specialists, so our SEO lead can hand a fix to the developers down the hall, and on our own builds the search decisions get made during content modeling instead of after launch.
Two other things that don't change. Every analytics property, Search Console account, and platform login stays in your name, and you see what we see, because reporting that only flatters the agency isn't reporting. And when the honest answer is that no amount of optimization will fix what's wrong, and the content model or the site architecture is the actual problem, that's the answer you'll get, even when it points at a smaller engagement than the one you asked about.
Let's have a chat →
Questions we get asked before an SEO engagement starts.
Is SEO dead? +
No, though the version of it that was mostly about ranking for a keyword list is winding down. Total search volume is still growing, and Google still sends more traffic to most sites than every assistant combined. What's changed is that a growing share of queries get answered without a click, so a program measured only in sessions will look like it's failing while it's actually working. We report visibility, citations, and conversions separately for exactly that reason.
What's the difference between SEO, AEO, and GEO? +
They describe three surfaces rather than three disciplines. SEO is the work of ranking in traditional results. AEO, or answer engine optimization, is the work of being cited in AI answers, including Google's AI Overviews. GEO, or generative engine optimization, is the work of being named and recommended inside generative responses from assistants like ChatGPT and Perplexity. The technical foundation underneath all three is largely the same, which is why we run them as one program. The differences are in prompt research, content formatting, crawler access, and how you measure it.
Can you guarantee we'll show up in ChatGPT? +
No, and nobody can. There's no submission form, no ranking dial, and no paid placement to buy. What we can do is make your content structurally eligible to be quoted, build the authority signals that models weight, benchmark where you stand across a set of target prompts, and then show you that measurement moving over time. Any agency guaranteeing a specific assistant result is describing something they don't control.
How do you measure AI visibility when there's no click? +
We track it as its own metric rather than trying to force it into a traffic number. That means monitoring your share of voice across the 25 to 50 prompts we establish at baseline, watching how the assistants describe your brand and your experts over time, and separately tracking referral sessions from ChatGPT, Perplexity, Gemini, and Claude in GA4, which do show up as their own channel. A citation with no click still shaped someone's shortlist, and the prompt tracking is how we can tell you it happened.
Should we block the AI crawlers or let them in? +
For most organizations, let them in, because a model that can't crawl you can't cite you. The exceptions are real: if your content is your product, if you license it, or if you're in a category where competitors are training on your material, blocking specific bots may be the right call. The decision should be explicit either way, and it usually isn't. Our team will show you what your robots.txt currently says about GPTBot, ClaudeBot, PerplexityBot, Google-Extended, Applebot-Extended, and CCBot, and what each of those permissions actually gets you.
Do we need an llms.txt file? +
It's cheap to add and worth having, and it isn't the thing that will change your visibility. An llms.txt file gives assistants a clean, structured map of what your site covers and which pages matter, and adoption across the model providers is still uneven. We implement it as part of the technical foundation and we won't oversell what it does on its own. The schema, the content structure, and the crawler permissions carry far more weight.
How long before we see results? +
Technical fixes and schema can show up in weeks, particularly on a site with an existing crawl relationship. Content and authority work generally takes six to twelve months to compound, because that's the pace at which these systems re-evaluate a source. Our projections reflect that, so the first quarter usually looks flat while the corrective work lands, and the gains build across the rest of the year. We'll tell you which bucket each recommendation falls into rather than implying they pay off at the same speed.
Do you write the content, or just tell us what to write? +
Either, and we'll scope whichever fits your team. Some clients want implementation-ready briefs (primary and secondary keywords, search intent, a competitive read on what's currently ranking, a recommended outline with headings, internal links in and out, and the E-E-A-T requirements) and their own writers take it from there. Others want us to handle it end to end, including writing, editing, optimization, and publishing support. The briefs are the same either way.
What happens to our rankings when we replatform? +
They hold if the redirect strategy is built during discovery, and they don't if it gets handled the week before launch. Our SEO team runs a full URL audit at the start, maps every existing URL to its destination, and preserves the internal linking and structured data that earned those rankings. After launch we watch crawl behavior, indexation, and position for a defined window rather than assuming it went fine. A dip of a few weeks is normal, and a permanent loss is a redirect failure, which is avoidable.
Can you work in Spanish or other languages? +
Yes. We run multilingual organic work with high-end strategic partners who handle the language side, while our team owns the technical foundation, the research, and the measurement. The important caveat is that translated content rarely performs, because search behavior differs by language rather than mapping cleanly onto the English keyword set, so the keyword and prompt research gets done natively in each market and the content gets adapted rather than run through translation. On the technical side, hreflang and canonical handling have to be right or your two language versions end up competing with each other for the same audience, which is the most common failure we inherit.
We're hiring you for a redesign, not for SEO. Does any of this apply? +
Yes, and it's most of the reason redesigns lose traffic. The decisions that determine organic performance (URL structure, the redirect map, navigation depth, content modeling, and whether priority content renders server-side) all get made during discovery and development, and they're expensive to revisit afterward. Our SEO lead is staffed onto design and development engagements with a defined allocation in discovery and again in implementation, so those decisions get made deliberately. You don't need a retainer for that, and we'll tell you if we think you'd benefit from one after launch rather than assuming it.
Can you work with our in-house team or our existing developers? +
Yes, and a lot of our organic work runs that way. Our SEO lead writes implementation-ready tickets with acceptance criteria, sits in your sprint rituals where that helps, and QAs the work once it ships, which is a different arrangement from handing over an audit and hoping. Where your developers are the ones implementing, the main thing we ask for is a staging environment and a way to see what changed, because verifying a fix landed correctly is part of the job.
How do you scope and price this? +
The audit and baseline phase is a fixed fee with a defined deliverable, so you know what it costs and what you get before it starts. Ongoing work is a monthly retainer sized to the hours the program needs, and what drives that number is site size, how much of the implementation your team can absorb, and whether content production is in scope. Pricing the audit separately means you can see the roadmap before committing to the program, and you can take that roadmap elsewhere if you'd rather.
Organic search work usually leads somewhere more specific.
If your traffic is slipping, or the assistants are describing you wrong, tell us what you're seeing.
Whether that's a full program across all three surfaces or a single audit that establishes where you stand, we'd like to hear it. We'll tell you what we'd do about it, what it would take, and whether we're the right people for the job.