Search is no longer about links

Technical writers are watching the same shift everyone else is watching: search engines are turning into AI-driven answer engines. Instead of sending users to web pages or help sites, these systems increasingly synthesize and deliver answers directly inside search results pages and chat interfaces.

For publishers who depend on referral traffic 📈, this raises alarms. For technical writers, it raises a different — and potentially more interesting — question:

If AI systems are now the primary readers of our content, are we writing in a way machines can actually understand?

Uncomfortable Truth: AI Can’t Make Sense of Unstructured Content Blobs

Large language models perform best when they can identify intent, scope, constraints, and relationships. Unstructured prose forces them to guess. That guessing shows up as hallucinations, incomplete answers, or advice taken out of context.

👉🏾 PDFs, long HTML pages, and loosely structured markdown were designed for human reading, not machine reasoning.

They obscure:

• What task the user is trying to complete

• Which product version applies

• What prerequisites or warnings matter

• Whether an instruction is conceptual guidance or procedural truth

When AI search engines ingest this kind of content, they flatten it. The result they provide may sound fluent, but it’s dangerously lacking in the precision department.

Read more

In AI and Accountability, Sarah O’Keefe argues that organizations are chasing the wrong goal with generative AI in content creation. Many businesses describe success as producing “instant free content,” but that’s a flawed metric. The true organizational goal isn’t volume of output — it’s content that supports business objectives and that users actually use.

AI Produces Commodity Content

O’Keefe points out that:

• Content is already treated as a commodity in many marketing teams

• Generative AI excels at producing generic, average content quickly

• That’s fine for low-level output (but it doesn’t create quality, accurate, domain-specific content on its own)

• When the goal is high-value content, simply doing “more” with AI fails

O’Keefe mirrors broader industry cautions that AI is very good at pattern synthesis but struggles with original, accurate, context-aware creation without structured inputs.

AI Can Raise Average Quality — But It Won’t Raise Above Average

O’Keefe emphasizes that:

• If your current content is below average, AI might improve it

• If your content needs to be above average, generative AI does not reliably get you there without expert oversight

• This performance gap stems from how AI models generate outputs — they produce the statistical average of what they’ve been trained on

This is an important distinction for tech writers whose docs must be precise, technically correct, and trustworthy.

Accountability Doesn’t Go Away With AI

A central theme is: authors remain accountable for the accuracy, completeness, and trustworthiness of content they publish, regardless of whether AI assisted in generating a draft. O’Keefe gives practical examples:

• AI can article-spin but still generate bogus reference

• AI won’t justify legal arguments or identify critical edge cases

• Errors from AI are still your responsibility if you publish the content

This has real implications for technical documentation where accuracy and liability matter.

The Hype Cycle Isn’t Strategic

O’Keefe warns that focusing internally on “how heavily we’re using AI” is a hype-driven metric, not a strategic one. Tech writers and content leaders should reframe the conversation:

• From “how much AI are we using?”

• To “how efficient and reliable is our content production process?”

• To “how good is the content we deliver for its intended use?”

This shift helps teams build accountability into their workflows instead of chasing novelty.

What This Means for Tech Writers

For tech writers specifically, here are the distilled takeaways:

1. AI should be a tool — not the goal: Use AI to handle repetitive, well-defined tasks, while keeping humans responsible for correctness and intent

2. Focus on quality over quantity: Audiences value clarity and accuracy more than volume

3. Maintain author accountability: Even when AI suggests or drafts content, the author still owns the final product’s correctness and accuracy, especially for technical and regulated content

4. Measure impact, not usage: Shift performance evaluation from AI adoption rates to content effectiveness and user outcomes

👉🏾 Read the article. 🤠

The Content Wrangler and Promptitude invite you to participate in our 2026 State of AI in Technical Documentation Survey, a five-minute multiple-choice survey that aims to collect data on where AI appears in documentation workflows, which tasks it supports, how often it is used, and the challenges that limit broader adoption.

We also seek to understand organizational support, tooling gaps, and where documentation professionals believe AI could deliver the most value in the future.

After the survey closes, we will analyze the results and publish a summary report highlighting the trends, patterns, and shared challenges across the industry. The goal is to provide documentation leaders and practitioners with credible data to guide decisions, advocate for better tools and training, and set realistic expectations for AI’s role in technical documentation.

Complete the survey and provide a valid email address to receive a copy of the final survey results report via email shortly after the survey closes.

The Content Wrangler is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.

It’s not too late! I hope you'll join me in Brussels at Hotel Le Plaza Brussels, February 2-3, 2026, for DITA EUROPE 2026 — the conference for technical writers and information developers looking to build advanced information management and delivery capabilities. If your technical documentation team produces content using the Darwin Information Typing Architecture (DITA) — or you’re considering doing so‚ this is the conference for you.

👉🏼 Grab your tickets now! Save 10% off the cost of registration using discount code — TCW — at checkout.

I’ll be delivering a talk with Dominik Wever, CEO of Promptitude — Powering AI and Virtual Humans with DITA Content — during which we’ll show how we built an AI-powered bot that delivers trustworthy, contextually accurate answers using technical documentation authored in DITA XML. Then, we’ll take it a step further—layering an interactive video agent (a virtual human) on top of the bot to create an engaging, human-like interface for delivering content.

Attendees will learn why structured content is essential for powering intelligent systems, how to responsibly connect DITA to AI models, and what opportunities virtual humans offer for technical communication. Whether you’re curious about future-ready documentation practices or looking for practical ways to extend the value of your content, this talk will offer both inspiration and concrete lessons.

DITA Europe is brought to you by the Center for Information Development Management (CIDM) — a membership organization that connects content professionals with one another to share insights on trends, best practices, and innovations in the industry.

CIDM offers networking opportunities, hosts conferences, leads roundtable discussions, and publishes newsletters to support continuous learning. CIDM is managed by Comtech Services, a consulting firm specializing in helping organizations design, create, and deliver user-focused information products using modern standards and technologies.

Day One: February 2, 2026

A large portion of Day 1 focuses on how AI actually behaves when it meets real-world documentation systems and legacy content — warts and all.

Key focus includes:

• AI benefits and failure modes in real-world use

• Aligning AI output with style guides and editorial standards

• Enhancing AI results through better context, tools, and editing workflows

• Deploying documentation directly to AI systems (not just to humans)

February 2 sessions emphasize that AI quality is constrained by content structure, governance, and determinability:

• AI in the Real-World: For Better and Worse — Lief Erickson

• Win Your Budget! — Nolwenn Kerzreho

• AI + DocOps = SmartDocs — Vaijayanti Nerkar

• DITA in Action: Strategies from Real-World Implementations — Maura Moran

• Determinability and Probability in the Content Creation Process — Torsten Machert

• From Binders to Breakthroughs: Cisco Tech Comms’ DITA Journey — Anton Mezentsev & Manny Set

• Enhancing AI Intelligence: Leveraging Tools for Contextual Editing — Radu Coravu

• From SMB’s to Enterprises: Real-World Content Strategy Journeys — Pieterjan Vandenweghe and Yves Barbion

• Aligning AI Content Generation with Your Styleguide — Alexandru Jitianu

• A Transition From One CCMS to Another – Lessons Learned — Eva Nauerth and Peter Shepton

• Cutting the Manual Labor: Scaling Localized Software Videos — Wouter Maagdenberg

• Revolutionizing Docs: DITA, Automation & AI in Action – Akanksha Gupta

• Future of Documentation Portals: From DevPortal to Context Management System — Kristof Van Tomme

• How AI is Boosting Established DITA Localization Practices — Dominique Trouche

• Trust Over Templates: Skills That Make A Content Transformation Work — Amandine Huchard

• Powering AI and Virtual Humans with DITA Content — Scott Abel

• DITA 2.0 + Specialization Implementation at ALSTOM — Thomas Roere

• Deploying Docs to AI — Patrick Bosek

• I’d Like Mayo With That: Our Special Sauce for DTDs — Kris Eberlein

Day Two: February 3, 2026

• Transforming Customer Support Through Intelligent Systems — Sharmila Das

• DITA and Enterprise AI — Joe Gollner

• Agentic AI: How Content Powers Process Automation — Fabrice Lacroix

• Why DITA in 2026: The AI Age of Documentation — Deepak Ruhil and Ravi Ramesh

• Content Automation 101: What to Automate and What Not To — Dipo Ajose-Coker

• Empowering DITA with Standards for AI in Safety-Critical Areas — Martina Schmidt and Harald Stadlbauer

• Taxonomy For Good: Revisiting Oz With Taxonomy-Driven Documentation — Eliot Kimber and Scott Hudson

• Upgrading the DITA Specification: A Case Study in Adopting 2.0 — Robert Anderson

• Hey DITA, Talk To Me! How iiRDS Graphs Infer Answers — Mark Schubert

• The Secret Life of Technical Writers: Struggles, Sins, and Strategies — Justyna Hietala

• Better Information Retrieval in Bringing DITA, SKOS, iiRDS Together — Harald Stadlbauer, Eliot Kimber, Mark Schubert

• If They Don’t Understand, They Can’t Act: Comprehension in DITA — Jonatan Lundin

• DITA, iiRDS, and AI Team Up for Smart Content Delivery — Marion Knebel, Gaspard Bébié-Valérian

• Pretty As a Screenshot: Agile Docs Powered by QA Tools — Eloise Fromager

• Bridging the Gap – Connect DITA, Multimodal Data via iiRDS — Helmut Nagy, Harald Stadlbauer

• Guardrails for the Future: AI Governance and Intelligent Information Delivery — Amit Siddhartha

• Highly Configurable Content using DITA, Knowledge Graphs and CPQ-Tool — Karsten Schrempp

• Building a Content Transmogrifier: Turning Reviews into Something Better — Scott Hudson, Eliot Kimber

👉🏼 Grab your tickets now! Save 10% off the cost of registration using discount code — TCW — at checkout.

The Content Wrangler is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.

Search used to be about matching keywords. Today, AI-powered search engines promise to “understand what users mean.” In practice, they struggle with something far more basic: intent.

Intent is the goal behind a search query—the job the user is trying to get done at that moment. Are they trying to learn, do, fix, compare, or verify? Intent answers why the question was asked, not just what words were typed.

For technical content, intent matters more than phrasing. A query like “configure OAuth token refresh” can signal at least four different needs:

• a conceptual explanation,

• a step-by-step task,

• a troubleshooting scenario,

• or a confirmation that a system even supports the feature.

The words don’t change. The intent does.

Why AI Search Engines Struggle With Intent

AI search systems are trained on language patterns, not user goals. That creates several hard problems:

Read more

The Darwin Information Typing Architecture (DITA) is an XML-based standard for creating, managing, and publishing technical documentation as modular, reusable content. OASIS maintains the DITA standard and provides access to information about its usage.

DITA matters because it forces clarity and discipline at the point of creation. Each topic has a single intent — concept, task, or reference — which makes content easier for people to scan and easier for machines to process. That structure enables reliable reuse, reduces duplication, and improves consistency across products, versions, and channels. Writers stop rewriting the same explanation five times and start managing one trusted source of truth.

The Content Wrangler is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.

DITA is also built for scale and change. Its specialization and inheritance model lets organizations extend the standard without breaking it, while maps assemble topics into many deliverables from the same content set. The result is faster updates, cleaner localization workflows, and publishing pipelines that can target PDFs, HTML, portals, and help systems from the same content base.

Finally, DITA future-proofs documentation for AI-driven delivery. Because topics are semantically typed and structurally consistent, search engines, chat systems, and retrieval-augmented generation pipelines can identify intent and return precise answers instead of dumping entire pages. In an era where documentation feeds intelligent systems, DITA gives technical writers a way to make content both human-readable and machine-ready.

DITA content is best managed in a component content management system designed specifically to help technical writers create, manage, and deliver documentation to the humans and machines that require it. 🤠

If you work with the Darwin Information Typing Architecture (DITA) — or plan to in the future — you likely to hear the word topic constantly. It sounds obvious; until you try to explain it to someone outside the DITA world, or worse, to someone who thinks a topic is just “a page” or “a section.” It isn’t.

A DITA topic is a small, self-contained unit of information designed to answer one specific user need. Not a chapter. Not a document. Not a dumping ground for everything related to a feature. One purpose. One job.

The Content Wrangler is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.

A DITA topic has three defining characteristics

1. It has clear intent

Every DITA topic exists to do exactly one thing: explain a concept, describe a task, or provide reference information. This is not an accident. DITA enforces intent through information typing—most commonly concept, task, and reference.

Related: Understanding DITA Topic Structure

That constraint forces writers to decide why the content exists before writing it. Is the user trying to understand something? Do something? Look something up?

📌 If you cannot answer that question, you do not yet have a topic.

This is one of the reasons DITA content works so well with search, delivery platforms, and AI systems. Intent is explicit, not implied.

2. It Stands on Its Own

A DITA topic is designed to be meaningful outside the context of a book or manual. Someone might encounter it through search, a chatbot, an embedded help panel, or an AI answer engine. Regardless of how they encounter it the topic must still make sense.

That means no “as described above,” no unexplained acronyms, and no reliance on surrounding pages or section headers to fill in critical gaps. Context travels with the topic through metadata, structure, and reuse — not through proximity.

This independence is what enables reuse at scale without copy-and-paste chaos.

3. It’s Structurally Predictable

A DITA topic follows a consistent structural pattern: a title, a body, and clearly defined elements inside. Tasks have steps. References have tables and properties. Concepts explain ideas, not procedures.

That predictability is not just for writers; it is for machines (computers, specifically). Structured topics allow systems to classify, filter, assemble, personalize, and deliver content dynamically. This is why DITA topics are increasingly valuable in AI-driven environments.

Large language models do not “understand” documents. They perform far better with small, well-typed, semantically clear units of information.

Why This Matters Now

As search engines turn into answer engines and documentation gets consumed in fragments, the DITA topic becomes the atomic unit of knowledge delivery. Writers who understand how to design strong topics are not just explaining products; they are engineering intent-aware, machine-processable content that feeds knowledge systems.

If you want AI solutions to give users the right answer instead of a plausible one, start with better topics.

If your content:

• clearly states its purpose,

• uses consistent information types,

• exposes relationships through structure and metadata,

AI can reason over it. If it doesn’t, AI guesses.

Structured, semantically augmented DITA content allows AI systems to:

• distinguish learning intent from execution intent,

• select the right topic type for the moment,

• avoid blending procedural steps into conceptual explanations,

• reduce hallucination caused by context collapse.

In short, it helps AI choose instead of invent.

What This Means For Tech Writing

Technical writers are no longer just explaining products. They are designing intent-aware knowledge systems.

Every time you:

• choose a topic type,

• enforce information typing discipline,

• add meaningful metadata,

• break monolithic docs into purpose-driven components,

you make it easier for AI search engines to deliver the right answer, not just an answer.

In an AI search world, visibility depends less on keywords and more on whether machines can understand what your content is for.

DITA doesn’t just support that future. It anticipates it. 🤠

The Content Wrangler is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.

Technical writers live in a strange middle ground. We translate what systems produce into something people can understand and use. Lately, that middle ground has gotten more crowded. Data teams, AI teams, and platform teams increasingly want documentation to behave like data—structured, atomized, and machine-friendly.

The assumption is that machines thrive on raw inputs and that humans can be served later by whatever the system generates.

That assumption is wrong!

The uncomfortable truth is this: data without context doesn’t just fail people—it fails machines too. The failure simply shows up later, downstream, disguised as “AI output.”

Most documentation problems trace back to a quiet but persistent confusion between four related — but very different — concepts: data, content, information, and knowledge.

They are often used interchangeably in meetings and strategy decks, as if they were synonyms. They are not.

Each represents a different stage in the creation, interpretation, and application of meaning. When those distinctions blur, teams ship docs that look complete but fail in practice (leaving both users and AI systems to guess at what was never clearly expressed).

A simple example, repurposed from Rahel Anne Bailie’s self-paced workshop for the Conversational Design Institute, Mastering Content Structure for LLMs, makes this painfully clear.

A Data Example

Imagine encountering the number 242.

On its own, it is nothing more than a value. Humans can’t reliably interpret it, and neither can an AI system. It could be a temperature, an identifier, a page number, or something else entirely. There is no intent encoded in it. No audience implied. No action suggested. It is easy to store and transmit, but useless for understanding.

Read more

By Mark Wentowski

Content pros are encountering new challenges in the Large Language Model (LLM) and agent era, highlighting the need for a different approach to authoring tools. Trust in LLM-generated content is the next hurdle, and transparency about the LLM’s information sources is needed.

Underscoring this, LLMs’ answers are only as good as the source content, which, in most cases, will be proprietary and will not be part of an LLM’s general intelligence about non-proprietary sources like blog posts. So the LLM needs to be an expert on your product, and someone needs to author this content, either with or without AI assistance.

The Current Paradigm

Right now, most authoring is aimed at publishing content for people. While chatbots can use this information to answer questions, the content isn’t designed with agents in mind. Humans remain an important audience, but we also need source content that works well for both LLMs and people. Similar to how web design shifted to a “mobile-first” approach, we could move to an “agent-first” design for documentation. This doesn’t mean humans are less important; it just means that focusing on agents is more challenging but will still meet human needs.

Related: AI-Powered Authoring: Will Machines Replace Technical Writers?

The Shifting Role of Authors

Authors should always retain the freedom to “just type something” in an editor, but moving forward, a bit more discipline will be required. As LLMs and agents become more adept at generating content (especially when supplied with verifiable facts) the role of the human author is evolving. Their primary responsibility will shift toward establishing a foundation for content verification. This includes building taxonomies, which might be generated automatically from source material and then refined by the author, or created manually for maximum control by a discerning analyst. Additionally, authors will help map product concepts to ontologies to enable deeper and more meaningful interconnections.

Taxonomies: How to Find Things

A taxonomy organizes items into categories within a clear hierarchical structure, much like a tree with branches and sub-branches. Each item is assigned to a category, making it easier to find by moving from broad groupings to more specific ones. These relationships typically take the form of parent-child or sibling connections.

We see taxonomies every day—for example, website navigation menus that let us drill down from broad categories to more specific items.

Why do taxonomies matter?

According to Earley Information Science:

Related: What Is the Difference between Taxonomy and Ontology?

Read more

If you’re a technical writer who has been told to “start using AI,” there’s a good chance this instruction arrived the same way most bad ideas do: casually, optimistically, and with absolutely no adjustment to your workload.

You were already juggling three releases, a backlog of outdated content, an SME who responds exclusively in emojis, and a publishing system held together by hope and duct tape. Now you’re also supposed to “use AI” — presumably in the five spare minutes between standup and your next fire drill.

When AI fails to deliver miracles under these conditions, leadership often concludes that AI is overrated. This is unfair to you, and frankly, unfair to the robot.

AI Isn’t a Magic Button — It’s a Process Multiplier

AI does not save time by default. It reallocates time.

To use AI responsibly in technical documentation work, you need time to:

• Experiment with real content, not toy examples

• Learn which tasks benefit from AI and which ones absolutely do not

• Validate output, because “confidently wrong” is still wrong

• Adjust workflows, prompts, review steps, and quality checks

If none of that time exists, AI use collapses into exactly one activity: drafting text faster.

Drafting is fine. Drafting is useful. Drafting alone is not transformation.

What Actually Happens Under Time Pressure

When documentation teams operate under constant time pressure, AI adoption tends to play out in a very predictable way. Writers reach for AI to generate first drafts because it offers the quickest visible win and creates the feeling of progress. The problem is that speed at the front of the process often pushes effort to the back.

Review cycles grow longer as someone has to verify facts, check tone, and untangle confident-sounding mistakes. Quality concerns begin to surface, not because AI is inherently flawed, but because it is being asked to compensate for a lack of time, structure, and planning.

As these issues accumulate, trust in the tool starts to erode. Conversations quietly shift from “How can we use this better?” to “This is risky,” and then to “Let’s limit where this can be used.”

Almost no one pauses to ask the more uncomfortable question: whether the real problem is that writers were expected to redesign workflows, adopt new practices, and safeguard quality without being given the time to do any of that work properly.

Instead, AI becomes the convenient scapegoat. Workloads stay the same, pressure remains high, and the system continues exactly as it did before—just with one more abandoned initiative added to the pile.

Why This Isn’t a Personal Failure

If you feel like you “don’t have time for AI,” that does not mean you are behind, resistant, or doing it wrong. It means your organization has not yet made room for change.

AI adoption requires slack in the system. Not endless slack. Just enough breathing room to think, test, and improve without breaking production.

Without that space, AI becomes another demand layered onto an already impossible job. And no technology (no matter how hyped) can fix that.

The Real Conversation We Need to Be Having

The question is not, “Why aren’t writers using AI more?”

The question is, “What work could we stop doing, simplify, or delay so writers can safely integrate AI into how documentation actually gets produced?”

Until that question is answered, AI will remain a side experiment — interesting, occasionally helpful, and permanently underpowered.

And the robot will continue to be blamed for a problem it did not create.

During the 2025 The LavaCon Content Strategy Conference the folks at Heretto gave away some cool merch (hats, t-shirts, and sweatshirts ). I’m pictured here with the “content” t-shirt 👕 — but in retrospect, I should have grabbed a sweatshirt —> because ❄️winter🥶 is here!

Now it’s your chance to score a “content” sweatshirt

Here’s how: Complete the Heretto Structured Content Survey (2 minutes max) and you’ll qualify to be entered into the drawing.

Complete the survey before December 31, 2025 to qualify. Winners to be contacted in January 2026.

Complete the Survey

The Content Wrangler is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.

Organizations everywhere are rethinking how they deliver customer support, and the smartest teams are turning to AI, structured content, and stronger cross-functional collaboration to do it. Earlier this year, Heretto released the State of Customer Self-Service Report, a data-rich look at how leading companies are transforming self-service into a strategic advantage. If you haven’t downloaded it yet, now is the time:

This report doesn’t just highlight trends. It reveals what top performers actually do to create self-service experiences that reduce effort, deflect tickets, and give customers the answers they need the moment they need them.

Download the State of Customer Self-Service Report now and see exactly where your organization excels, where your gaps are, and what steps you can take to improve.

Inside The State of Customer Self-Service Report

You’ll learn how organizations are using:

• AI to power smarter, more personalized support. The report explores how teams are adopting AI responsibly, pairing machine intelligence with strong content foundations to improve relevance and accuracy across the customer journey.

• Structured content to move beyond scattered, inconsistent knowledge bases. High performers are investing in reusable, well-governed content that scales across channels and markets — and makes AI far more reliable.

• Cross-functional collaboration to break down silos between product, support, documentation, and marketing. Teams that treat content as shared infrastructure, rather than as a departmental afterthought, deliver better self-service outcomes and faster ROI.

Why This Matters

Self-service is now the first stop for a growing majority of customers. That shift raises expectations for clarity, findability, consistency, and personalization. What you’ll see in the report is that organizations who meet those expectations aren’t relying on guesswork; they’re aligning strategy, structure, and technology to deliver high-quality content at scale.

The report also explores the barriers teams run into: limited staffing, siloed tooling, unclear ownership, and inconsistent processes. More importantly, it outlines how leaders overcome those challenges and build momentum.

Next Step: Benchmark Your Content Against Industry Realities

If your organization delivers customer-facing content (technical docs, support articles, onboarding flows, or product guidance) you should download this report and see how your strategy compares to the findings.

Use it as a benchmark. Use it as a wake-up call. Use it as the evidence you need to advocate for smarter, more scalable content operations.

The insights are available now. Your competitors may already be acting on them. Don’t wait!

Download the State of Customer Self-Service Report now and see exactly where your organization excels, where your gaps are, and what steps you can take to improve. 🤠

The Content Wrangler is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.

Everyone wants AI. Few leaders want to think about how it actually works. If you’ve been told to “use AI” without being given the time, structure, or systems needed to make that request anything other than comedy, you already understand the tension behind this whole conversation.

Technical writers are being asked to conjure miracles from the literary equivalent of a junk drawer. Leadership wants smarter chatbots. They do not want to hear that the chatbot cannot perform surgery on a heap of inconsistent topics last updated during the dot-com bubble.

So here we are: expected to make AI sing while working with content that can barely hum.

Why “Unsexy” Structural Work Determines AI Success

This is why structural investments matter. Not “nice-to-have someday” investments — the foundational work that determines whether any AI initiative lives, dies, or produces the kind of answers that make support teams question their career choices. When you bring this up, you may be met with the blank stare reserved for people who talk about metadata in public. That’s fine. Your job isn’t to dazzle them. Your job is to translate structural needs into something leaders recognize as business value, not writerly fussiness.

Start with a shared definition. Structural investment isn’t mysterious. It’s the unglamorous backbone of any AI program: structured content, consistent terminology, clean versioning, healthy repositories, sensible taxonomies, and governance that doesn’t rely on six people remembering to “just ping each other.”

These aren’t luxuries. They’re the conditions that determine whether AI can find what it needs, recognize what it sees, and deliver answers that don’t read like refrigerator poetry.

Why AI Falls Apart Without Structure

You already know why these foundations matter. AI needs patterns. It needs clarity. It needs content that behaves predictably. Hand an AI a pile of duplicative, contradictory, unlabeled topics and it will react like any reasonable being: it will guess. And it will guess poorly. When terminology is inconsistent, retrieval collapses. When topics are half-updated, summarization breaks. When metadata is missing, routing becomes improvisational theater. None of this is AI’s fault. The systems simply can’t work when the content isn’t prepared to support them.

But here’s the real trick: when you advocate for these fixes, don’t talk to leadership in the language of writers. They don’t wake up in the morning worrying about chunking strategies or rogue taxonomies. They think in terms of efficiency, risk, and repeatability. They care about faster onboarding. Lower support costs. Reduced rework. Improved customer experience. These are all things structural improvements deliver — they just don’t sound like they’re about topic types.

So translate the need. “We need metadata” becomes “AI tools cost more to maintain and produce less reliable output when content isn’t findable.” “We need better structure” becomes “Support tickets rise because AI can’t distinguish which answer is current.”

Shift the conversation away from what you want and toward what the business needs — because they’re the same thing, even if the vocabulary differs.

Start Small, Show The Pain, And Make The Case

Advocacy works best when you stop trying to fix everything at once. Start with a small audit of the pain everyone already knows about: duplicated content, outdated topics, inconsistent terminology, missing metadata, chaotic versioning. These problems are not secrets; most organizations simply lack the language to connect them to AI performance. You can make the connection clear. Collect examples of AI outputs that went sideways because of structural issues. Show the time your team loses rewriting machine-generated answers that pulled from outdated content. Nothing builds support faster than seeing the cost of inaction.

Next, outline the smallest possible set of structural improvements that would produce measurable gains. You don’t need a five-year transformation plan. You need a believable starting point.

Propose a contained, achievable fix — a pilot for cleaning key topics, a simple metadata model, or a tidy workflow that reduces version confusion. Package the request in a one-page summary that explains the problem, the underlying cause, the small investment required, and the business outcome. Keep it concise enough that someone could read it between meetings, but clear enough to make the value unmistakable.

Find Allies Who Need This As Much As You Do

You’ll also want allies. Not the kind who nod politely, but the ones who actually suffer alongside you: support leads who hate inconsistent answers, engineers who want fewer interruptions, localization teams who want stable content, product managers who want AI-ready documentation without having to understand how it works. When they see how structure helps them, they will become your best advocates.

When the time comes to demonstrate return on investment (ROI), stay grounded in metrics that matter. Show how reuse improves. How duplication drops. How AI outputs require fewer edits. How search becomes more accurate. How support volume shifts when answers get consistent. You’re not just proving that structure works — you’re showing how it delivers the outcomes AI promised but couldn’t reach alone.

Lead The Charge: AI Works When Writers Demand What It Needs

AI cannot function without the foundations writers have championed for years. And writers cannot keep pretending that “working harder” is a substitute for the structural investment AI requires. If your organization wants AI that performs with reliability, explainability, and consistency, you are uniquely positioned to guide them there.

Leading the charge for these investments isn’t self-serving. It’s responsible stewardship of the content the entire business depends on. The message you want to leave with leadership — and with yourself — is straightforward: AI succeeds when structure succeeds. And structure succeeds when technical writers stop quietly compensating for the gaps and start leading the conversation about what good actually requires. 🤠

Join us for Coffee and Content with Jack Molisani, founder of ProSpring Technical Staffing and the organizer of the LavaCon Conference on Content Strategy and Technical Documentation Management. On this episode we’ll unpack the findings of the 2025 Technical Documentation Staffing Survey from The Content Wrangler and ProSpring — the most comprehensive look yet at how content teams are hiring, where they’re struggling, and what skills are most in demand.

Survey data show that while over 60% of managers face headcount authorization hurdles, many are also frustrated by a lack of candidates with structured authoring, API documentation, and AI/prompt engineering experience.

See: Call for presenters —> LavaCon 2026 — Charlotte, North Carolina

The Content Wrangler is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.

We’ll explore how these realities are reshaping hiring priorities — and what it means for writers, recruiters, and managers navigating today’s competitive job market.

Expect an honest discussion about:

• Which roles are most in demand (and which are fading)

• Why compensation and remote work remain hot-button issues

• How AI and automation are changing the skills employers value

• What organizations can do to attract and retain top technical talent

Whether you’re hiring, job hunting, or simply curious where the profession is headed, this conversation offers a rare data-driven snapshot of the industry — straight from one of its leading staffing experts.

Can’t make it? Register, and we’ll send you a link you can use to watch a recording when it’s convenient for you. Each 60-minute episode of Coffee and Content is recorded live. On-demand recordings are made available shortly after the live show ends.

Coffee and Content is brought to you by The Content Wrangler and is sponsored by Heretto, a powerful component content management system (CCMS) platform to deploy help and API documentation in a single portal designed to delight your customers.

The Content Wrangler is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.