Titles, meta descriptions, headings and URLs — the on-page fundamentals that still move the needle.
On-page SEO is the part of the discipline you fully control. You cannot make another site link to you and you cannot make an engine cite you, but you decide every title, heading, URL, and word on your own pages — which makes on-page the most controllable lever available and the one worth getting right before anything harder. The fundamentals have barely changed in a decade. What has changed is that clean on-page structure now doubles as preparation for being cited by AI engines, so the same work pays twice.
On-page work covers everything on the page itself: the title, the meta description, the heading structure, the URL, the body content, the images and their alternative text, and the internal links out. These elements tell engines what the page is about and tell users whether it is worth their attention, which makes them foundational rather than incidental. Every other discipline — authority, technical health, content strategy — ultimately acts through a page that has to be intelligible in itself.
The reason to prioritise it is control. Building authority takes months and depends on other people; technical fixes often depend on engineering capacity; but on-page improvements can usually be made immediately by the person who owns the content. This makes on-page the highest-certainty work available, and it is why a page that underperforms should be checked for on-page problems before more expensive explanations are pursued. Understanding on-page as your most controllable lever is why it belongs first in most improvement sequences.
The title is the single most important on-page element. It is a primary signal of what the page is about, it is what appears as the clickable headline in results, and it is frequently what a person uses to decide whether to click. A good title states clearly and specifically what the page offers, includes the terms the page genuinely targets, and reads naturally to a human rather than as an assembly of keywords.
Practical guidance is unglamorous but effective: put the most important terms early, keep it within a length that will display rather than truncate, make each title on your site distinct so pages do not appear interchangeable, and describe the page honestly so that clicks are not immediately regretted. A title that overpromises earns a click and loses the visitor. Understanding the title’s dual role — signal to engines, promise to readers — is why it deserves more care per word than anything else on the page.
The meta description does not directly determine ranking, but it frequently determines whether a listing is clicked, which makes it worth writing deliberately rather than leaving to be auto-generated. Its job is to expand on the title’s promise and give a reader enough to decide, ideally by stating what they will get from the page rather than describing the page in the abstract.
Engines often rewrite descriptions to match the specific query, which leads some to conclude they are not worth writing. The better reading is that a well-written description is used when it fits and replaced when the engine can do better for a particular query — so writing one costs little and sometimes helps. Keep it within a displayable length, make it specific, and avoid duplicating it across pages. Understanding its role as a click decision rather than a ranking factor is what keeps effort proportionate.
Headings organise a page for both readers and machines. The convention that works is a single H1 stating what the page is about, with H2s marking major sections and H3s marking subsections beneath them, forming a logical hierarchy rather than a set of visual size choices. Using heading levels for styling rather than structure breaks the hierarchy and removes the signal they exist to provide.
Beyond structure, headings should be descriptive: a heading that states what its section covers helps a reader scanning and helps an engine identify what the section addresses. Phrasing headings as the questions people actually ask strengthens this further, since it matches how queries and prompts are phrased. Understanding headings as structural signals rather than typography is why one H1, a logical hierarchy, and descriptive wording are worth enforcing consistently across a site.
URLs should be readable, descriptive, and stable. A URL containing plain words describing the page’s subject is intelligible to both people and engines, while one composed of identifiers and parameters conveys nothing. Keeping URLs reasonably short, using hyphens between words, staying lowercase, and reflecting the site’s logical structure are the conventions that produce clean addresses.
Stability matters more than most teams expect. Changing a URL breaks links pointing to it and discards accumulated signals unless redirects are handled carefully, so URLs are worth getting right at creation rather than optimising later. The practical rule is to choose a sensible, durable URL when the page is created and then leave it alone. Understanding URLs as both a modest signal and a long-term commitment is why they warrant a moment’s thought at publication and little attention thereafter.
You can’t make a site link to you or an engine cite you, but every on-page element is yours to set. Clean structure and genuine topic coverage serve rankings and extraction at the same time.
Images need alternative text describing what they show, which serves accessibility first and comprehension second: a screen reader uses it, and an engine uses it to understand content it cannot otherwise interpret. Descriptive alt text states what is actually in the image, plainly, rather than repeating keywords or padding with terms unrelated to the picture.
The related discipline is not placing important information only in images. A statistic, a comparison, or a key instruction that exists solely as a graphic is invisible to anything reading text, which means it cannot be indexed, extracted, or read aloud. The practical rule is that anything that matters should exist in text, with images supporting rather than carrying it. Understanding this is why image handling is both an accessibility requirement and a visibility one, with the same remedy serving both.
The single most consequential on-page judgment is whether the page matches the intent behind the queries it targets. A page can contain every relevant term and still fail because it is the wrong kind of page — a sales page for an informational query, a shallow overview where the searcher wanted a procedure. Matching intent means building the kind of content that would actually satisfy the person asking.
The obsolete alternative is keyword stuffing: repeating terms to signal relevance. It does not work, it reads badly, and it can trigger the systems designed to catch it. Modern engines understand meaning rather than counting occurrences, so covering a topic genuinely and thoroughly is what demonstrates relevance. Understanding the shift from term repetition to intent matching and topic coverage is the difference between on-page work that helps and on-page work that actively harms.
Engines have invested heavily in identifying content created primarily to rank rather than to help, and such content is systematically demoted. The characteristics are recognisable: pages covering topics the site has no genuine expertise in, content assembled by summarising what already ranks, answers padded to reach a length rather than to add substance, and material that leaves the reader no better informed than before.
The corrective is to write for the person who will read it, and to publish only where you have something genuine to contribute. This is a strategic constraint as much as an editorial one: it argues against covering topics simply because they have volume. Understanding that helpful-content systems penalise search-first writing is why on-page optimisation has to sit on top of content that would deserve to exist even if search did not, since optimisation cannot rescue material that has nothing to offer.
Clean on-page structure is what makes a passage liftable as well as rankable. DUNkē tracks whether you’re actually cited across eight AI engines — per prompt, against competitors — so you can see your on-page work translating into citations.
The on-page fundamentals turn out to align closely with what makes content citable by AI engines. Descriptive headings create the labelled boundaries retrieval uses to identify passages. A clear title and logical hierarchy make the page’s subject and organisation legible. Substantive, intent-matched content gives an engine something worth drawing on. Text-based information ensures the substance is readable at all.
This convergence is genuinely useful, because it means the same work serves both objectives rather than requiring separate programmes. The additional disciplines that GEO adds — answer-first opening statements and self-contained passages, covered in their own pieces — layer naturally onto sound on-page structure. Understanding that clean on-page doubles as GEO is why on-page work has become more valuable rather than less, and why it remains the sensible first investment.
The links a page makes outward are part of its on-page profile. Linking to genuinely relevant related pages helps readers continue, distributes authority through your site, and signals relationships between your content that help engines understand how your topics connect. Descriptive anchor text on those links tells both readers and engines what to expect at the destination.
The discipline is relevance and restraint: links placed where they genuinely help, with anchors describing the target honestly, rather than a dense mesh of links inserted mechanically. Excessive linking dilutes the signal each link carries and degrades readability. Understanding outward links as part of on-page work connects this discipline to site architecture, and it is why linking decisions belong in the writing process rather than in a later optimisation pass.
A practical audit runs a short sequence. Does the title state clearly and specifically what the page offers? Is there exactly one H1, with a logical heading hierarchy beneath it? Are headings descriptive, ideally question-shaped? Is the URL readable and stable? Do images carry meaningful alt text, with no important information trapped in graphics? Does the content match the intent behind its target queries and cover the topic properly?
Each failure maps to a specific, quick correction, which is what makes on-page auditing efficient work compared with most visibility improvements. Running this sequence across your most important pages usually surfaces several straightforward fixes with immediate effect. Understanding the audit as a repeatable checklist is why on-page improvement can be delegated and scaled once the sequence is written down, rather than depending on individual judgment each time.
The recurring failures are consistent across sites. Duplicate or auto-generated titles make pages look interchangeable. Multiple H1s or heading levels chosen for size break the hierarchy. Keyword-stuffed titles and copy read badly and risk demotion. Content that mismatches intent fails regardless of optimisation. Important information locked in images is invisible. And publishing on topics where the site has no genuine expertise triggers exactly the systems designed to catch search-first content.
The remedies follow directly: write distinct honest titles, enforce one H1 and a logical hierarchy, cover topics genuinely rather than repeating terms, build the kind of page the intent requires, keep substance in text, and publish only where you have something to contribute. Understanding these failure modes is practically useful because most on-page underperformance is a matter of these errors rather than of missing sophistication, and each is correctable in minutes.
Beyond containing the right terms, a title has to persuade someone scanning a page of similar-looking options. What works is specificity: a title promising something concrete outperforms one describing a topic generally, because the reader can tell what they will get. Numbers, defined scope, and stated outcomes all help, provided they are honest about the content behind them.
What consistently fails is the opposite of specificity — vague titles that could describe any page on the subject, and inflated titles promising more than the page delivers. The second is worse, because it earns the click and immediately loses the visitor, which signals dissatisfaction rather than success. The practical test is whether someone reading only your title would correctly predict what they find. Understanding titles as promises rather than labels is why the honest specific version outperforms both the vague and the overstated.
There is no target word count that improves rankings, and treating length as a goal produces padded pages that read badly and perform worse. What correlates with performance is completeness — fully answering the question, including the follow-ups and edge cases a serious treatment requires — which happens to produce longer pages when the subject genuinely demands it.
The distinction matters practically because it changes the editing instinct. Writing to a word count encourages adding; writing to completeness encourages covering what is missing and cutting what is not needed. A thorough page on a simple question can be short and still be complete. Understanding that length is a symptom rather than a lever is why the useful question during editing is whether anything a reader needs is absent, not whether the page has reached some number.
Some queries carry mixed intent — people searching a product category may want to understand options, compare them, or buy immediately — which raises the question of whether one page can serve all three. It sometimes can, if structured so each intent is addressed in a clearly-signposted section rather than blended into undifferentiated prose.
Where the intents are genuinely divergent, separate pages usually serve better, with clear linking between them so a visitor can move to what they actually wanted. The failure mode is a page hedging between purposes and satisfying none, which is common on commercial pages trying to educate and sell simultaneously. Understanding how to handle mixed intent is why examining what currently ranks is informative — the results reveal which interpretation the engine considers dominant.
Most on-page value in an established site comes from improving existing pages rather than optimising new ones, because the archive is where the accumulated impressions and authority sit. A systematic pass — checking titles, hierarchy, intent match, and structure across the pages that already receive impressions — typically surfaces quick corrections with immediate effect.
Prioritisation follows performance data: pages receiving impressions but few clicks often have title or intent problems; pages ranking just below the visible positions often need better topic coverage. Both are cheap fixes relative to producing new content. Understanding that on-page work is largely retrofitting is why an early audit of existing high-impression pages is usually the highest-return starting point in an established site.
On-page quality that varies page by page produces an inconsistent site, and consistency itself carries weight — both in how engines assess overall quality and in how efficiently a team can work. Establishing conventions for title patterns, heading structure, URL format, and required elements makes quality reproducible rather than dependent on whoever wrote each page.
The practical mechanism is a documented standard plus checks at publication, so conventions are enforced rather than merely recommended. This also makes onboarding faster and reduces the audit burden later, since fewer defects ship. Understanding consistency as an operational property is why on-page standards belong in the publishing process, and why the highest-leverage on-page work in a growing site is often standardisation rather than individual page improvement.
Several once-common practices have stopped mattering and some now cause harm. Keyword density targets are obsolete, since engines assess meaning rather than counting occurrences. Meta keywords tags are ignored entirely. Exact-match repetition in headings reads badly and risks demotion. And bolding terms for emphasis toward engines rather than readers achieves nothing.
Knowing what to stop doing is useful because these practices persist in checklists and consume effort that produces no benefit while sometimes producing risk. The general principle behind their obsolescence is consistent: techniques that signal relevance to a machine without serving a reader have been systematically devalued. Understanding what no longer matters is why on-page work has converged on simply making pages clear, honest, and genuinely useful.
On-page work sits underneath everything else in the discipline, which is worth making explicit. Technical work determines whether a page can be reached and read; on-page determines whether what is read is intelligible and relevant. Authority determines whether the source is trusted; on-page determines whether the specific page deserves to represent that source for a given query. Content strategy determines what pages should exist; on-page determines whether each one does its job.
This positioning explains why on-page problems are so costly relative to how cheap they are to fix: a page with strong authority behind it and sound technical foundations can still fail because its title misrepresents it or its structure buries the answer. The practical implication is to verify on-page soundness before investing in anything more expensive, since it is the layer where effort produces results fastest and where problems most often hide in plain sight.
The most efficient place to enforce on-page quality is at publication, since correcting a page later costs more than getting it right once. A brief pre-publication check — title specific and honest, single H1 with logical descriptive hierarchy, URL clean and final, alt text present, intent matched, nothing important trapped in images — catches nearly everything.
Encoding this as a required step, ideally with automated checks for the mechanical items, removes a whole category of recurring audit findings. It also prevents the accumulation that makes later audits so daunting, where hundreds of pages each carry small defects. Understanding publication as the cheapest point of enforcement is why on-page standards belong in the workflow rather than in a periodic remediation project.
On-page SEO covers the elements you fully control — titles, meta descriptions, headings, URLs, images, content, and outward links — which makes it the most certain lever available and the right place to start when a page underperforms. The fundamentals are stable: a specific honest title, a single H1 with a logical descriptive hierarchy, a readable stable URL, meaningful alt text, and substance kept in text rather than trapped in graphics.
What matters most is matching the intent behind the queries you target and covering the topic genuinely, since modern engines understand meaning rather than counting terms, and helpful-content systems actively demote material written for search rather than for readers. And because descriptive headings, clear structure, and substantive text are exactly what makes a passage retrievable, clean on-page work now doubles as preparation for AI citation — the same effort serving both rankings and the answer layer.
“On-page is the only part of this you fully control. Get the title, the hierarchy, and the substance right, and the same work that ranks the page is what makes it quotable.” The Age’X Research Team
Get a free GEO audit — the same analysis behind every article here.