Content strategy for AI search: a teardown of one agency's total restructuring

Most agencies talk about content strategy for AI search in the abstract — new buzzwords bolted onto the same audit template, same deliverables, same org chart. iPullRank published something more useful: an account of actually rebuilding its service model, team structure, and workflow around AI search, with enough specificity to argue with. That's rare, and it's worth taking apart piece by piece rather than nodding along.

The piece is iPullRank's account of its own restructuring, published July 16, 2026, and authored by Francine Monahan. It describes an agency-wide shift away from a single service tier toward three, the creation of a named discipline called Relevance Engineering, and a new six-stage engagement sequence that replaces the old one-off-audit-then-retainer model. No client counts, no before/after traffic numbers, no revenue lift — the piece is a structural account, not a results deck. That's actually the most useful thing about it: you can evaluate the restructuring on its own terms instead of chasing an unverifiable stat.

What iPullRank changed in its content strategy for AI search

Before the restructuring, iPullRank sold one flavor of engagement regardless of client size or maturity. After it, there are three: Emerging, for founder-led companies and startups building an AI search foundation from zero; Growth, for mid-market organizations that already have an SEO program and a content team but a gap between strategy and execution; and Elite, for enterprise accounts running multiple business units, distributed teams, and regulated categories. That's a fairly standard agency move — tiered service models are as old as agencies — but the reason it matters here is what triggered it. The piece describes the same failure mode across client sizes: traffic disappearing in ways that are hard to explain in a report, attribution getting confusing, and AI platforms recommending competitors in categories where the client had spent years building authority. A single audit template couldn't diagnose that consistently across a five-person startup and a regulated enterprise brand, so the tiering exists to scope the diagnostic work to the client's actual exposure, not just their budget.

The bigger change sits underneath the tiers: a new named practice called Relevance Engineering. iPullRank defines it as a technical discipline built specifically around how AI systems access, retrieve, and interpret content — query fan-out, passage-level performance, retrieval paths, and how a brand shows up in model outputs. That's a deliberate break from calling this work "SEO with an AI angle." Classic technical SEO audits crawl a site and check indexation, site speed, and structured data against a search engine's crawler behavior. Relevance Engineering, as described, is checking whether a retrieval system can find and correctly interpret a specific passage of content when a model fans a query out into a dozen sub-queries — a different failure surface entirely, and one most technical SEO audits still don't test for.

Worth being precise about what query fan-out actually means, since the term does real work here. When a user asks an AI system a question, the system frequently breaks that single query into several related sub-queries before it retrieves anything — clarifying scope, checking definitions, pulling comparison data — then synthesizes an answer from whatever passages each sub-query returns. A page can rank first for the parent query and never get retrieved for the three sub-queries that actually determine whether it ends up in the answer. That's the mechanism a Relevance Audit is built to test, and it's also the mechanism most legacy SEO tooling has no visibility into at all, because rank trackers were built to watch a single query, not a fan-out tree.

The six-stage workflow behind the new content strategy for AI search

The workflow is the part worth studying closest, because it's the part that determines what actually happens on an account, week to week. It runs Strategic Planning, Content Engineering, Solutions Engineering, Conversation Engineering, Creative, then Measurement Engineering. Read as a sequence rather than a list, it looks like a real production line, not a rebrand of a single deliverable.

Strategic Planning is standard account management with new paperwork names — a Joint Business Plan is a shared goals document, and giving it a formal name is a repositioning move, not a new capability. Content Engineering is where the real diagnostic work sits: a Relevance Audit is functionally a new artifact, checking whether a page's passages surface correctly when a model fans a query into sub-queries, which a keyword-rank audit was never built to measure. Solutions Engineering is the closest analog to classic technical SEO, but retargeted — instead of crawl budget and Core Web Vitals, it's asking whether an LLM's retrieval system can actually reach and parse a given page. Conversation Engineering is link building and digital PR with a new name, aimed specifically at the domains that show up as citations in AI answers rather than the domains that pass the most link equity. Creative is content production, now billed on a credits model instead of a monthly retainer of X articles — a pricing change, not a workflow change. Measurement Engineering is reporting, rebuilt around AI visibility signals like citation share and passage retrieval instead of purely organic rank tracking.

One thing to flag before moving to what's genuinely new here: the source piece never quantifies any of this. No stated lift in citation share after moving a client through the new sequence, no case study attached to a specific tier, no timeline for how long a Relevance Audit takes to complete. That's not a knock — plenty of good methodology write-ups are process documents, not results decks, and a company describing its own restructuring has every incentive to round toward flattering numbers if it has them, so the absence is arguably more credible than a suspiciously clean stat would be. But it does mean you should read the workflow as a structural template to test against your own accounts, not as a proven system with a track record attached.

What's substantive vs. what's repackaged SEO

Take a position on this rather than treating every stage as equally new, because they aren't. Two things in this restructuring are genuinely new mechanics: the Relevance Audit as a diagnostic artifact, and Solutions Engineering as a technical practice retargeted at retrieval systems instead of crawlers. Both require a skill set most content and SEO teams don't currently have on staff — someone who understands how retrieval-augmented generation actually selects and ranks passages, not just how a search index does. That's a real hiring or training gap, and it's the part of this restructuring that should change what you actually do, not just what you call it.

That skill gap shows up first in job descriptions, and it's worth being concrete about it. A content strategist hired to run a Relevance Audit needs to understand how retrieval-augmented generation selects and re-ranks passages, how embedding-based similarity differs from keyword matching, and how to read a citation-share report the same way they'd read a rank tracker. Almost nobody currently on a content team was hired against that spec, because it didn't exist eighteen months ago. Closing it is either a training investment inside an existing content marketing function or a targeted hire — and either path takes longer than adopting a new label for an old audit, which is exactly why this is the half of the restructuring that actually matters.

Everything else is competent repositioning, and there's nothing wrong with that — agencies reposition constantly, and giving a familiar service a sharper name is how you sell it to a buyer who's currently panicking about AI search. Three pricing tiers is standard agency segmentation. Conversation Engineering is link building and digital PR, retargeted at citation sources instead of link equity — a smart targeting shift, not a new discipline. Omnimedia Content Audits are cross-channel content audits, which good agencies were already doing before anyone said "AI search." A prompt library for content production is a documentation habit, not a technical breakthrough. None of that is a criticism of iPullRank — the piece doesn't oversell any of it as revolutionary, and the framing that "the most expensive thing you can do in AI search right now is use the wrong strategy" is the right warning regardless of which half of the restructuring you're looking at. A mislabeled diagnostic doesn't hurt you. A missing one does.

“The most expensive thing you can do in AI Search right now is use the wrong strategy.”

That line is the one worth sitting with. It's not about picking the wrong tool or skipping a tactic — it's about misdiagnosing what kind of problem you have. A team that runs a standard technical SEO audit against an AI-visibility problem will produce a clean report and fix nothing, because it's testing for the wrong failure mode. That's the actual argument for something like our generative engine optimization work: the audit has to be built for retrieval systems specifically, not adapted from a crawler-based checklist after the fact.

Regulated categories make the stakes concrete. iPullRank's Elite tier is scoped to enterprise brands with multiple business units and regulated categories specifically because those clients can't afford the diagnostic to be wrong — a mislabeled or off-brand claim surfacing in an AI answer isn't just a lost click, it's a compliance problem with a much longer cleanup cycle than a dropped ranking. That's a real reason the wrong-strategy warning lands harder at enterprise scale than it does for a startup still building its first content library, and it's a reasonable justification for tiering the diagnostic work by exposure rather than selling everyone the same audit regardless of what's actually at risk.

What a smaller team should adopt first

You don't need six named stages, three pricing tiers, or a rebrand to get value out of this. Most teams — in-house or agency — should take three things from it, in order. First, separate your diagnostic work: run one audit for classic search ranking factors and a distinct pass for AI retrieval and citation behavior, because they test for different failures and a single combined checklist will miss the one it wasn't built for. Second, start a prompt library and a content engineering brief template now, even in a spreadsheet — the discipline of writing down what language and structure a piece needs to be retrievable is more valuable than the tool you use to store it. Third, rebuild your measurement layer before you rebuild anything else. If your reporting still only tracks organic rank and sessions, you're flying blind on the exact failure mode the source article describes — traffic disappearing in ways that are hard to explain in a report is a measurement problem before it's a content problem.

Sequence matters more than headcount here. A single content lead can run a Relevance Audit on the ten highest-traffic pages in under two weeks if the diagnostic already exists — the bottleneck was never people, it was not knowing what to test for. Once that audit exists, use its output to prioritize the next three months of work: pages with strong classic rankings but weak retrieval signals get fixed first, because they're the highest-confidence wins — the authority is already there, and the fix is almost always structural rather than a full rewrite. Pages with neither signal go to the back of the queue until the foundation exists to support them.

Skip the parts that only make sense at agency scale. You don't need three formal service tiers if you're a single in-house content team — that's a sales structure, not an operating one. You don't need a Joint Strategic Activation Plan as a named document; you need the actual planning conversation it represents. And resist the urge to rename your existing content audit process just to sound current — a genuinely useful audit for AI search is defined by what it tests, not what it's called. The naming exercise is marketing work. Do it after the diagnostic actually works, not instead of building it.

We've watched this same split play out with clients directly: teams that rebuilt their measurement and diagnostic layer first, before touching org structure or service naming, got to a defensible answer about their AI visibility gap faster than teams that started with a rebrand. That's the same sequencing behind how we approached content restructuring in a recent client engagement — diagnose what's actually broken at the retrieval level before you touch the org chart or the pricing page.

Do this next: pull your last quarter of organic traffic and citation data side by side, and flag every page where the two diverge — rank stable, citations or AI-referral traffic down. That divergence is your Relevance Audit, whether or not you ever call it that. Fix the retrieval problem on those pages before you write a single new piece of content, and before you decide whether your team needs a new org chart at all.

See where you are cited today

A free snapshot audit of your rankings and AI citations before we ever talk.