On-page SEO is everything you can change inside your own website: content, titles, page structure, internal links, loading speed. Off-page SEO is everything that happens outside it: the links other sites give you, mentions of your name, reviews, citations in the places where your sector gets talked about.
So far, everyone says the same. What most guides on this topic don't say is in which order to work on them, and how to tell which of the two is holding you back right now. That is the part that decides whether an SEO project produces anything in the first six months or burns budget on the right activities done at the wrong time.
The difference in six rows
| On-page SEO | Off-page SEO | |
|---|---|---|
| What it covers | Content, titles and headings, URL structure, internal links, images, structured data, reading experience | Inbound links, unlinked brand mentions, reviews, Google Business Profile listing, citations in directories and publications |
| Who decides | You. Nobody has to give you permission | Other people. You can earn it or ask for it, not decide it |
| Where the work shows | In the page's code and text | On other people's websites |
| How it is measured | Position and clicks for the individual page, query coverage | New referring domains, mentions of the name, share of visibility on the cluster |
| What happens if you get it wrong | The page doesn't capture demand: it stays indexed and invisible | In the worst case you carry a link profile Google considers manipulative |
| How long it takes to fix | An afternoon, for a single page | Months, and partly out of your hands |
The last row is the one nobody usually looks at, and it matters most when you have to decide where to start.
What on-page SEO really covers
The lists you find online put the page text and an image's alt text on the same level. They don't carry the same weight. The on-page work that moves things is nearly always one thing: making what the page says match what the person was looking for when they typed that query. Everything else is finishing, useful but secondary. If you haven't yet checked what kind of answer searchers expect, start from search intent and keyword research, not from tags.
On titles, one thing is worth knowing that hardly anyone mentions: what you write is often not what appears in the results. Cyrus Shepard of Zyppy analysed 80,959 title tags across 2,370 sites in 2022 and found that Google had rewritten 61.6% of them. The share rises to 99.9% for titles over 70 characters and falls to 39-42% in the 51 to 60 range. The practical conclusion: writing a long meta title stuffed with keywords is pointless, because two times out of three it gets replaced.
The same goes for the meta description. Google Search Central's documentation is explicit: snippets "are primarily created from the page content itself", and the meta description is used only when it describes the page better than the extracted text. It isn't a ranking factor, it is a persuasion element that Google sometimes decides not to use.
The most underrated part of on-page work is internal links. They are the only authority signal you can assign yourself, without asking anyone for anything, and on most of the sites we analyse they are left to chance or confined to a "related articles" block nobody reads. The full list of elements is in the on-page SEO guide.
What off-page SEO covers
For years off-page SEO meant link building and nothing else. Today it covers at least four different things: backlinks, unlinked mentions of your name, reviews, and the listings where others check that you exist, starting with Google Business Profile.
On links, Google says two things worth keeping apart. The first is about discovering pages: its documentation on how Search works says that some pages are discovered "when Google extracts a link from a known page to a new page". The second is about abuse, and it is the definition in the spam policies, updated on 28 August 2026: "Link spam is the practice of creating links to or from a site primarily for the purpose of manipulating search rankings." Buying and selling links, excessive link exchanges, sponsored articles with optimised anchors and links in the footers of third-party sites are among the examples listed.
How much do links really weigh? The clearest number I know comes from Ahrefs' study of about 14 billion pages, published in December 2023: their index contains about 20 million pages with no referring domains at all, and of these only 2,997 get more than a thousand search visits a month. One in 6,671. It is a correlation, not proof of cause. But it gives a sense of how rare it is to get far without anyone linking to you.
Technical SEO is not the third pillar
Many guides on this subject add technical SEO as a third pillar, with the same standing as the other two. In practice it doesn't work like that, and it is a split that confuses whoever has to set priorities.
Technical SEO is a prerequisite, not an area of ongoing investment. If Google can't crawl the site, or pages take six seconds to load on mobile, nothing you write counts. But once those problems are fixed, they stay fixed: nobody gains rankings by publishing a sitemap.xml every month. Technical work is concentrated at the start and during migrations, then it becomes maintenance. On-page and off-page, on the other hand, never end.
Which order to do them in, and how to tell where you are stuck
The right order isn't a general rule, it is a diagnosis. You make it by opening the Performance report in Google Search Console and looking at two columns, impressions and average position, for the queries you care about. Three cases, three different answers.
- Impressions close to zero. Google doesn't consider you relevant for those searches, or a dedicated page simply doesn't exist. It is an on-page problem through and through, and no link will fix it. You need a page that answers that question.
- Impressions present, position between 8 and 20. You are relevant but not convincing enough. Here on-page work still pays well: expanding the page, covering the sub-questions, fixing internal links pointing to it. This is the band where you see the fastest movement.
- Stable position between 4 and 10 on contested commercial queries. You have done what you could on your own. The pages ahead of you nearly always have more referring domains, and from here on the lever is off-page.
The order matters because the two levers work on different timescales, and getting the sequence wrong costs months. Again from Ahrefs, on a sample of over a million URLs analysed between September and October 2023: only 1.74% of newly published pages reach the top 10 within a year. Of those that make it, 40.82% get there within the first month. Put another way: if a page doesn't move in the first few weeks, it is almost never a matter of waiting, it is a matter of rewriting it. And if you don't know where to start looking, a well-run SEO audit starts from exactly this distinction.
The data behind this reasoning
All the figures quoted above, with the primary source and the market measured. The global studies apply to any market with caution; the Italian rows are there because they frame the case study that follows.
| Figure | Value | Source and sample | Market |
|---|---|---|---|
| Pages that get no traffic from Google | 96.55% | Ahrefs, index of about 14 billion pages, December 2023 | Global |
| Pages with no referring domains that exceed 1,000 visits a month | 2,997 out of about 20 million (1 in 6,671) | Ahrefs, same study | Global |
| Titles rewritten by Google in the results | 61.6% | Zyppy, 80,959 titles across 2,370 sites, 2022 | Global |
| New pages that reach the top 10 within a year | 1.74% | Ahrefs, over 1 million URLs, September-October 2023 | Global |
| Correlation between web brand mentions and visibility in AI Overviews | 0.656 (Spearman) | Ahrefs, 75,000 brands, December 2025 | Global |
| Correlation between Domain Rating and visibility in AI Overviews | 0.326 (Spearman) | Ahrefs, same study | Global |
| Italian firms with at least 10 employees that have a website | 76.5% | Istat, Businesses and ICT, December 2025 | Italy |
| Italian firms that sell online | 14.7% | Istat, same report | Italy |
The two Istat rows say a lot about the Italian market our case study comes from. Three firms in four have a website, fewer than one in seven use it to sell. Most of those sites don't have a link problem: they have a problem of pages that don't answer any question.
A case from our work: from position 88 to position 2
Macropix is a Milan-based manufacturer of custom LED screens. When the project started in 2020, the site didn't appear for any of its sector's category keywords on Google Italy. The figure we quote most often isn't the first place for "ledwall", a keyword with 4,400 searches a month in Italy: it is the row for "monitor pubblicitario" (advertising display), which moved from position 88 to position 2. Position 88 is Google's ninth page, meaning zero traffic.
That jump didn't come from a link building campaign. It came from a dedicated page, well written, on a subject the company genuinely had something to say about: on-page work, entirely. In July 2026, on the LED wall cluster, macropix.it has a 25.55% share of voice, ahead of amazon.it, which stops at 18.08%. That part, the ability to stay ahead of a domain with incomparably higher authority, is where five years of accumulated off-page signals weigh in: trade fairs, sector publications, installations cited by others. The full numbers are in the piece on SEO for manufacturers, and similar projects are collected on the projects page.
Our reading of it is this: on-page gets you into the game, off-page decides how long you stay in it.
What changes now that some answers are written by a model
This is the new element that most guides on this topic haven't yet taken on board, and it shifts the relative weight of the two areas.
In December 2025 Ahrefs published a study of 75,000 brands, by Louise Linehan, measuring the Spearman correlation between various factors and visibility in answers generated by ChatGPT, Google AI Mode and AI Overviews. Web brand mentions correlate at 0.664 with ChatGPT, 0.709 with AI Mode and 0.656 with AI Overviews. Domain Rating, the metric that sums up the strength of a link profile, stops at 0.266, 0.285 and 0.326 respectively. YouTube mentions reach 0.737 and are the factor with the highest correlation in the whole study.
A correlation is not a ranking factor. But the direction is clear: being named counts for more than being linked. For a typical small or mid-sized company this makes things easier, not harder, because a mention in a trade publication or a trade show press release can be obtained without negotiating a link. If visibility in AI answers is what you are specifically after, we cover it in the guide to generative engine optimisation and in the dedicated service.
Three mistakes we see almost every time
- Buying links before you have pages. It is the wrong order in the most expensive way possible. The links point to pages that don't answer any query, and Google classes them as what they are, according to the definition of link spam it has put in writing.
- Treating on-page as a checklist of tags. Alt text, keyword in the title, density: these are finishing touches on a page that works, not a way to make a page with nothing to say work.
- Measuring off-page by counting links instead of mentions. Given the correlation data above, a report that only lists referring domains is looking at half the picture.
If you need the split between the two areas to decide where next year's budget goes, we work through the full reasoning case by case in our SEO projects, and the off-page side has its own guide in off-page SEO.
Frequently asked questions
In almost every case with on-page, because it is the only one you control directly and because without relevant pages links have nowhere to pass value. The exception is a site that is already well structured and stuck between fourth and tenth position on contested commercial queries: there the remaining lever is off-page.
No. It also covers unlinked mentions of your name, reviews, citations in directories and trade publications, and the Google Business Profile listing. In the Ahrefs study of 75,000 brands from December 2025, web brand mentions correlate with visibility in AI answers more strongly than Domain Rating does.
It is a prerequisite, not a third area of ongoing investment. It is concentrated at the start of the project and during migrations: once crawling, indexing and speed problems are fixed, they stay fixed. On-page and off-page, by contrast, need constant work.
It happens, but it is rare on contested queries. The Ahrefs index analysed in December 2023 contains about 20 million pages with no referring domains, and only 2,997 of them exceed a thousand search visits a month. On very specific niche keywords the odds are higher, because the competition is weaker.
On the on-page side, movement shows early: in the Ahrefs study of over a million URLs, 40.82% of the pages that reach the top 10 get there within the first month. On the off-page side you think in quarters, because it depends on when others decide to link to you or mention you.
Sources cited
- Ahrefs, 96.55% of Content Gets No Traffic From Google, December 2023
- Ahrefs, How Long Does It Take to Rank in Google?, 2023
- Ahrefs, AI brand visibility correlations, December 2025
- Zyppy, Google Rewrites 61% of Page Title Tags, 2022
- Google Search Central, Spam policies for Google web search
- Google Search Central, In-depth guide to how Google Search works
- Google Search Central, Control your snippets in search results
- Istat, Imprese e ICT, anno 2025 (Businesses and ICT, 2025), December 2025
One thing the industry rarely says: of the two, the only one that is truly irreversible is off-page. A badly written page can be rewritten in an afternoon, and the following week Google has already reassessed it. A link profile built by buying packages stays attached to the domain for years and partly can't be undone, because those links aren't yours and nobody is going to remove them for you. Anyone choosing where to start would do well to weigh this too, not just which of the two levers promises to move rankings first.
