How-To Pages vs Tutorial Pages: What LLMs Actually Prefer
How-to pages and tutorial pages look similar on the surface and behave very differently in retrieval pipelines. A how-to page answers “how do I X” with a numbered procedure, a small set of prerequisites, and a clear success state. A tutorial page walks a reader through a worked example, with context, alternatives and educational framing. Both formats are useful; both are cited by AI engines, but for different query classes and under different conditions. Most B2B sites publish a confused hybrid that performs neither job well. This piece sets out which format wins which query class, what the retrieval pipelines look for in each, and how to choose between the two when planning a content programme.
The Two Formats, Sharply Defined
A how-to page is a procedural document. The user knows what they want to do and needs the minimum instruction to do it. The page should be scannable, short on context, heavy on numbered steps, and explicit about what success looks like at the end. The reader is not learning; they are executing. “How to verify domain ownership in Cloudflare.” “How to add a webhook in Stripe.” “How to migrate a Drupal site from version 9 to version 10.”
A tutorial page is a learning document. The user has a broader goal and needs to build understanding alongside execution. The page should explain why each step exists, name alternatives, and surface common failure modes. The reader expects to come away knowing more than just how to do the immediate task. “Setting up authentication in a Next.js application.” “Building your first MCP server.” “Migrating from REST to GraphQL: a worked example.”
The retrieval pipelines understand the difference because the user queries reveal it. “How to X” maps to how-to pages. “How does X work” or “X tutorial” maps to tutorials. A page that confuses the two intents (a how-to that meanders into education, a tutorial that omits context to look procedural) underperforms both.
What We See in Audits
On the Angular 17 fintech SPA audit, the firm’s documentation included both formats interleaved on the same URLs. Users searching for procedural answers (“how to configure Fastly cache headers for Angular”) were getting pages that included three paragraphs of CDN theory before reaching the actual configuration. The retrieval pipelines were skipping these pages in favour of community blog posts that put the answer in the first 200 words. The fix was format-aware splitting: the procedural answer moved to a how-to page, the theoretical context moved to a tutorial, and both linked to each other explicitly.
On the industrial-materials manufacturer’s content engine, the audit surfaced 727 DIY “installation” false positives in the contamination triage. The underlying confusion was format-level: trade pages that should have been written as tutorials for licensed installers were being interpreted as DIY how-to instructions by both users and the retrieval pipelines. Re-categorising the pages as tutorials, with explicit audience markers (“for licensed installers”), reduced misclassification and lifted citation share for the right query class.
On the BFSI content programme that shipped 794 briefs across four batches, each brief was assigned to a format type at the topic-clustering stage. Of the 794, roughly 380 were how-to format (“how to apply for X loan”, “how to check Y eligibility”) and roughly 250 were tutorial format (“understanding the X process”, “preparing for Y application”). The remaining 164 were comparison, definition or explainer content. The format-aware allocation was visible in the downstream Pydantic validation as a structural field; pages where the body did not match the declared format were flagged for rewrite.
The Architecture of a Citable How-To
Six properties show up consistently in how-to pages that earn AI citations.
The H1 names the task in user phrasing. “How to verify domain ownership in Cloudflare” beats “Domain verification: a comprehensive overview”. The H1 should match the query the user typed; deviation costs ranking and citation share.
Prerequisites listed in a small block. Three to seven items, each one a specific dependency. “You need a Cloudflare account, the domain added to your Cloudflare zone, and access to your domain registrar’s DNS settings.” This block prevents the page from failing on users who do not yet meet the prerequisites, and it gives the retrieval pipeline a clear extraction surface.
Numbered steps with imperative verbs. “1. Click X. 2. Enter Y. 3. Submit Z.” Not “you will then proceed to click X”. The imperative voice is shorter and matches the procedural intent. The numbered structure is what the retrieval pipeline preferentially lifts.
Each step’s success state described inline. “After step 4, you should see a green checkmark next to your domain.” The success state is what the user needs to verify progress. It is also a strong signal to the retrieval pipeline that the procedure is real and tested.
A “troubleshooting” section at the end. Three to five common failure modes, each with a specific remedy. This section is what wins citations on “X not working” queries, which sit immediately adjacent to “how to X” queries in the engine query log.
HowTo schema with step-by-step structure. schema.org/HowTo with HowToStep nodes. The schema reinforces the page’s procedural classification. The visible content must match the schema; mismatches are treated as cloaking and demoted.
The Architecture of a Citable Tutorial
Tutorials need a different structure to earn citations.
The H1 frames a learning goal. “Building authentication into a Next.js application” beats “How to add auth to Next.js”. The frame signals to both the reader and the retrieval pipeline that the page is educational, not purely procedural.
An “what you will learn” block in the first 100 words. Three to five learning outcomes. This is the block the retrieval pipeline extracts as the page’s intent summary.
Sectioned progression with named milestones. Each major section is a named milestone the reader passes through. “By the end of this section, you will have a working auth provider.” The milestones make the page navigable and citable section by section.
Context paragraphs between code blocks. Tutorials, unlike how-tos, explain why each step matters. The context paragraphs are short (60 to 120 words) and answer the “why” implicit in the step. These are the paragraphs that get cited on “how does X work” queries.
Named alternatives at decision points. “We are using NextAuth for this tutorial. The alternatives are Auth0 (commercial), Clerk (managed), and Supabase Auth (open-source plus hosted).” Named alternatives expand the query surface the page can be cited for.
Article or LearningResource schema. Not HowTo. The schema type matters because it tells the retrieval pipeline what kind of content this is. Schema-format mismatch is treated as a signal that the content is unreliable.
How-To vs Tutorial Decision Matrix
| Dimension | How-To | Tutorial |
|---|---|---|
| User intent | Execute a known task | Build understanding plus execute |
| Query phrasing | “how to X” | “X tutorial”, “how does X work” |
| Word count | 600 to 1,200 | 1,500 to 3,500 |
| Voice | Imperative (“Click X”) | Explanatory (“We click X because…”) |
| Schema | HowTo | Article or LearningResource |
| Citation context | Procedural answers | Educational explanations |
| Failure mode | Becoming a tutorial accidentally | Becoming a how-to accidentally |
Most B2B content fails at the boundary. Picking the format consciously, declaring it in schema, and writing to it produces measurable yield gain.
Which Format LLMs Cite for Which Query
The retrieval pipelines route queries to formats with predictable patterns.
How-to pages dominate procedural queries. “How to X” with a specific named task routes to how-to format, and the retrieval pipeline scores HowTo schema and numbered-step structure heavily. ChatGPT and Claude both quote step blocks; Perplexity often cites the full step list.
Tutorial pages dominate exploratory and educational queries. “How does X work”, “X explained”, “X for beginners” route to tutorial format. The retrieval pipeline weights conceptual context, named alternatives, and worked examples. Tutorials earn fewer citations per page than how-tos, but each citation tends to be longer and to bring more context.
The two formats cross over on hybrid queries like “set up X”. This phrasing is ambiguous: the user might want a procedural answer (how to do it) or an educational one (how to think about it). Pages that target this kind of query benefit from making the format choice explicit in the H1 (“How to set up X in 10 minutes” versus “Setting up X: a working tutorial”) and structuring the body accordingly.
The Anti-Patterns
Three patterns reliably hurt citation yield regardless of format.
The first is the screenshot-only how-to. A page that consists primarily of screenshots with minimal text is invisible to retrieval pipelines that do not OCR images. The screenshots should be supplementary; the procedural text should stand alone.
The second is the marketing-pivot tutorial. A tutorial that starts as education and pivots into a product pitch at step five loses trust prior. The retrieval pipeline reads the pivot as commercial intent disguised as educational content. Tutorials should keep the product mention minimal, late, and clearly separated.
The third is the unversioned procedure. Tools update. A how-to that references a UI element that has been renamed, or a CLI flag that has been deprecated, is worse than no how-to at all because it actively misleads users. Versioned tutorials with explicit version markers (“this tutorial is for X v3.2”) and a quarterly review cadence are the working pattern.
Practitioner Takeaway
- Audit your existing how-to and tutorial content for format consistency. Pages where the H1 promises one format and the body delivers another should be split or rewritten.
- Pick the format at brief-writing stage. Each content brief should declare “how-to” or “tutorial” as a structural field, and the writer should be held to it.
- Match schema to format. HowTo for how-tos, Article or LearningResource for tutorials. Schema-format mismatch is treated as a cloaking signal.
- Version your tutorials. Tool updates break tutorials. A version marker in the H1 or in a visible badge plus quarterly review prevents the page from misleading users.
- Cross-link how-tos and tutorials on the same topic. The user often needs both: the tutorial to understand, the how-to to execute. Internal links between the two compound the topical cluster’s authority.
How This Connects
How-to and tutorial pages are the post-decision surface in the commercial content stack. They handle the user who has already chosen the product (via the comparison page) and worked through the pricing (via the pricing page). The companion architectures sit in comparison pages, the new money page and pricing pages and LLM comparison queries. The wider retrieval logic that determines which format wins which query sits in how LLMs decide which sources to cite. For SaaS firms specifically, where the how-to and tutorial surfaces are usually the most-trafficked content, our SaaS growth engineering notes have the working playbook.
Frequently Asked Questions
Can one page be both a how-to and a tutorial?
Rarely well. The user intent at query time is usually one or the other. The working compromise is a short how-to section embedded inside a tutorial, with the tutorial’s H1 framing the page as educational and the how-to block clearly delineated.
Should tutorials include video?
Optional. Video helps human readers but is opaque to most retrieval pipelines unless transcribed. If video is included, a full transcript or a written walkthrough should accompany it for the pipeline to extract.
How long should a how-to page be?
600 to 1,200 words is the working range. Shorter than 400 words and the page may lack the prerequisites and troubleshooting that make it citable. Longer than 1,500 and the page starts to drift into tutorial territory.
Do screenshots help or hurt retrieval?
Help, if they accompany clear text descriptions. Hurt, if they replace the text. The retrieval pipeline cannot see the screenshot, so any procedural detail that exists only in an image is invisible.
What about troubleshooting pages as a separate format?
Troubleshooting pages are a specialised how-to format. They benefit from a problem-symptom-solution structure and from explicit error string mentions. The citation dynamics match how-to pages otherwise.
If your post-decision content is not earning the citation share your traffic suggests it should, the format mismatch between how-to and tutorial is usually a meaningful contributor. An audit can map the specific pages where the rewrite would pay back fastest.