Live tool
Internal Linking Simulator
Upload an export of your internal links to estimate PageRank, centrality and click depth. Audit your anchor text and simulate link changes based on real Google patents.
- Input
- CSV / XLSX link export
- Metrics
- PageRank, centrality, click depth
- Extras
- Anchor text audit, simulations
- Processing
- In your browser
The model
Watch PageRank flow through a site
PageRank is easier to understand as movement than as a formula. In the animation below, every page on a small demo site starts with the same score. On each step, a page shares most of its score between the pages it links to, and a small remainder is spread evenly across the whole site. Pages that collect links from many places grow whereas pages that only send links shrink.
You can change the damping factor, switch between the classic and position-weighted models and add or remove links to play around with the scores.
If you would rather not use the animation: the demo site has about 25 pages. After roughly 30 iterations the scores stop changing. The homepage ends up with the largest share since every page links to it from the navigation. The hub pages linked from that navigation come next, blog posts linked only from their category page sit near the average, a privacy page linked only from the footer scores well in the classic model but poorly once footer links are down-weighted, and a page with one inbound link sits at the bottom. A PDF with no outgoing links collects score and passes none on. The same data is available as a table inside the animation with the "Show as table" control.
Process
How it works
-
Export your internal links
In Screaming Frog, crawl the site and choose Bulk Export → Links → All Inlinks. In Sitebulb, export All Links from the Link Explorer. Any CSV, TSV or XLSX with a "from" and a "to" column will work.
-
Map columns & set weights
Columns are detected automatically and you can correct them. Choose how nofollow, duplicate and non-200 links are handled, and adjust how much a content, navigation or footer link is worth.
-
Explore, simulate, export
Sort pages by estimated PageRank, check click depth and anchor text, then add or remove links and compare before with after. Export every table as CSV or one XLSX, with the settings used.
What internal PageRank is, and why it still matters
PageRank is the idea that a link is a vote, and that votes from important pages count for more than votes from unimportant ones. Google stopped showing a public score years ago, but its representatives have repeatedly said that PageRank is still used internally, and the documents from the 2024 Content Warehouse leak refer to several PageRank variants.
You can't control who links to your site. You control every link inside it. Internal links decide which pages a crawler finds first, how often it returns, and how the authority your site has earned is shared between your pages. A product page that is five clicks deep with one inbound link is telling search engines it doesn't matter much, whatever the sitemap says.
Internal PageRank runs the PageRank calculation on your internal links alone. It ignores backlinks, so it isn't a prediction of rankings. It answers a narrower and more useful question: given the way this site links to itself, which pages does it treat as important? When the answer doesn't match your commercial priorities, you have an internal linking problem you can fix.
How the PageRank formula works
The original PageRank patent (opens in a new tab) describes a "random surfer" who clicks links forever. Most of the time the surfer follows a link on the current page, chosen at random. Occasionally they get bored and jump to a random page instead. A page's PageRank is the share of time the surfer spends on it.
PR(page) = (1 − d) / N + d × Σ ( PR(linking page) / links on that page )
dis the damping factor, the probability that the surfer follows a link. The usual value is 0.85.Nis the number of pages. The(1 − d) / Nterm is the random jump, shared equally by every page.- Each linking page splits its score equally between its outgoing links. A page with 100 links passes a hundredth of its score through each one.
- Pages with no outgoing links (PDFs, for example) would trap the surfer, so their score is redistributed across the whole site.
The scores depend on each other, so they are calculated by repetition: start every page at 1 / N, apply the formula, and repeat until the numbers stop moving. The scores always add up to 1.
A three-page worked example
Take a site with three pages. A links to B and C. B links to C. C links back to A. With d = 0.85, each page gets a baseline of 0.15 ÷ 3 = 0.05, plus 85% of what its linkers send.
| Page | Receives links from | Calculation | PageRank |
|---|---|---|---|
| A | C (its only link) | 0.05 + 0.85 × PR(C) | 0.388 |
| B | A (one of two links) | 0.05 + 0.85 × PR(A) ÷ 2 | 0.215 |
| C | A (one of two) and B (its only link) | 0.05 + 0.85 × (PR(A) ÷ 2 + PR(B)) | 0.397 |
Two things are worth noticing. C wins because it receives two links, including everything B has to give. And A is almost level with C despite having a single inbound link, because that link comes from the strongest page and isn't shared with anything else. Who links to a page, and how many other links sit beside that link, matter as much as the raw count.
The reasonable surfer: not every link is equal
The original model treats every link on a page as equally likely to be clicked. Nobody browses like that. Google's Reasonable Surfer patent (opens in a new tab) replaces the random surfer with one who is more likely to click some links than others. It lists the features that could feed that probability: where the link sits on the page, its font size and colour, the anchor text and the words around it, whether it is an image, and how well the source and target match topically.
The practical reading is that a link in the body of a relevant article probably passes more than the same link in a footer that appears on every page. The simulator's weighted model applies this by giving each link position a coefficient, and splitting a page's score in proportion to those weights instead of equally.
| Link position | Default weight |
|---|---|
| Content / body | 1.00 |
| Unknown | 0.70 |
| Aside / sidebar | 0.50 |
| Breadcrumb | 0.40 |
| Navigation / header | 0.30 |
| Footer | 0.15 |
These numbers are our assumptions. The patent describes the idea, not the values, and Google has never published them. That is why every coefficient is visible and editable, and why the classic unweighted score is always shown alongside. If a page looks strong in the classic model and weak in the weighted one, it is living on boilerplate links. That is worth knowing whichever model is closer to the truth.
Seed pages and distance
A later Google patent, covered in The SEO Papers' write-up of seed-based PageRank (opens in a new tab), ranks pages by how far they are from a set of trusted seed pages. Each link has a length, and a link from a page with many outgoing links is longer than a link from a focused page. The closer a page is to a seed, the better it scores.
On the open web the seeds are trusted sites. Inside one site, the natural seed is the homepage: it is usually the most linked-to page from outside. The simulator offers three seed-based metrics:
- Click depth: the fewest clicks needed to reach a page from the homepage.
- Seed distance: the shortest path from the nearest seed, where each link's length is
1 + log2(links on the source page). Two pages can both be three clicks deep, but the one reached through focused pages is closer than the one reached through three mega menus. - Personalised PageRank: the weighted model, with the random jump landing only on your seeds. Add the pages that earn most of your backlinks as seeds to approximate where authority enters the site.
Anchor text as a relevance signal
Links carry meaning as well as weight. Google's anchor text indexing patent (opens in a new tab) describes storing the text of each link, and text near it, against the target page, so that a page can be indexed for words that only appear in links pointing to it. Your internal anchors are a description of each page written by the one site you fully control.
The anchor report shows what that description says. For every page you get the full list of anchors it receives and how varied they are. Site-wide, you see how many links have empty or generic anchors ("click here", "read more"), and anchor conflicts: the same phrase pointing to two or more different pages. Conflicts are a common cause of the mixed signals behind keyword cannibalisation, so cross-check them with the Cannibalisation Analyser.
The diversity flags are informational. Exact-match internal anchors are normal and are not a penalty risk in the way manipulative external anchors can be. A page with 400 inbound links that all say "here" is simply a missed opportunity.
How to read each metric
| Metric | What it is | How to use it |
|---|---|---|
| PR multiple | Estimated PageRank × number of pages. 1.0 is the site average. | The default sort. Priority pages below 1.0 are under-linked. |
| PR score (0–10) | A log-scaled display of the same value. | A quick visual scale. It is not toolbar PageRank. |
| Classic vs weighted PR | The original equal-split model beside the position-weighted one. | A big gap means the page depends on navigation or footer links. |
| In-degree / unique inlinking pages | Links received, and how many different pages send them. | Find pages with very few referring pages. |
| Out-degree | Links on the page. | Pages with hundreds of links pass very little through each one. |
| Degree centrality | Links in and out, divided by the number of other pages. | Compare how connected pages are across sites of different sizes. |
| Click depth | Clicks from the homepage. | Important pages deeper than three clicks need a shorter route. |
| Seed distance | Weighted shortest path from the nearest seed. | Separates pages at the same depth by the quality of the route. |
| PR from nav share | Share of a page's incoming equity that arrives through navigation, header or footer links. | High values mean few editorial links point at the page. |
Internal linking fixes the simulator helps you find
Orphan and near-orphan pages
Upload a list of all your URLs (a sitemap export is enough) alongside the link file. Pages in the list that receive no links are orphans. Pages with one or two inbound links and a PR multiple near the floor are the next thing to fix. Link to them from a relevant hub or retire them.
Hub pages that hoard or waste equity
Category and pillar pages usually hold the most internal PageRank after the homepage. Check where they send it. A hub with 300 links spreads itself thin; a hub that links to twenty well-chosen pages lifts all of them. In Simulate mode, add links from your strongest hubs to an under-linked priority page and read the before and after.
Equity leaking to redirects, 404s and non-indexable pages
If your export includes status codes, the overview shows the percentage of estimated PageRank flowing to non-200 URLs. Links to redirects still work, but each one is a wasted hop, and links to 404s are a dead end. Updating the links at source is usually the cheapest internal linking win on a site. The Log File Analyser will tell you whether search engine bots are wasting crawl requests on the same URLs.
Bloated navigation and footers
Every link added to a site-wide menu dilutes every other link on every page. Use the pattern rule in Simulate mode (for example, "remove /blog/ from the footer on every page") to see what a slimmer template would do before you ask a developer to build it.
Anchor conflicts
Where one phrase points to several pages, choose the page that should rank for it and change the other links to point there, or reword them. The internal linking guide covers how to plan this across a whole site.
Limitations
- It is a model. The formulas come from published patents and graph theory. The weights are assumptions. Nothing here is Google's live calculation.
- No external links. Backlinks bring authority into specific pages, and the simulator can't see them unless you add those pages as seeds.
- Your crawl is the limit. Links injected by JavaScript are only included if your crawler rendered them. Pages the crawler never found are missing altogether.
- Position data varies by crawler. Where a link position column is missing, it is inferred from the link path, and the tool tells you how many links were inferred.
- PageRank is one signal among many. A well-linked page with thin content won't rank. Use the output to prioritise, then confirm changes against Search Console data.
Sources
- PageRank (opens in a new tab) (The SEO Papers) · US 6,285,999 B1, "Method for node ranking in a linked database" (opens in a new tab)
- Reasonable Surfer (opens in a new tab) (The SEO Papers) · US 7,716,225 B1, "Ranking documents based on user behavior and/or feature data" (opens in a new tab)
- Seed-based PageRank (opens in a new tab) (The SEO Papers) · US 9,165,040 B1, "Producing a ranking for pages using distances in a web-link graph" (opens in a new tab)
- Anchor text indexing (opens in a new tab) (The SEO Papers) · US 7,308,643 B1, "Anchor tag indexing in a web crawler system" (opens in a new tab)
Frequently asked questions
What is an internal linking simulator?
It is a tool that turns a list of your internal links into a graph, estimates how link equity (PageRank) is distributed across your pages, and then lets you test changes before you make them. Add a link, remove a footer link site-wide or delete a page, and the simulator recalculates every score and shows which pages gain and which lose.
Is this my real Google PageRank?
No. Nobody outside Google knows the values or weights Google uses today, and Google stopped publishing toolbar PageRank in 2016. Every number here is an estimate from a transparent model based on published patents and standard graph theory. Treat the scores as a way to compare pages on your own site and to compare a "before" with an "after", not as a measurement of what Google has stored.
What file do I need to upload?
A CSV, TSV or XLSX with one row per link and at least two columns: the page the link is on (from) and the page it points to (to). Screaming Frog's Bulk Export > Links > All Inlinks works with no column mapping. Exports from Sitebulb, JetOctopus and other crawlers work through the column-mapping step. Optional columns such as anchor text, link position, link path, follow and status code unlock position weighting, the anchor text audit and equity-leak reports.
Is my crawl data uploaded to a server?
No. The file is parsed and every metric is calculated in your browser, inside a Web Worker. Nothing is sent to techSEOstack or to any third party, and nothing is stored once you close the tab.
Why do navigation and footer links count for less than links in the content?
Google's Reasonable Surfer patent describes weighting each link by how likely a person is to click it, using features such as its position on the page. A link in the body text of a relevant article is more likely to be clicked than the fortieth link in a footer. The default coefficients (content 1.00, sidebar 0.50, breadcrumb 0.40, navigation 0.30, footer 0.15) are our assumptions, not Google's values. They are all editable, and you can switch to the classic unweighted model to see how much the weighting changes the picture.
What does "PR multiple" mean?
It is a page's estimated PageRank multiplied by the number of pages on the site. A multiple of 1.0 is exactly the average page, 3.0 is three times the average and 0.2 is a fifth of it. Raw PageRank values are tiny decimals that shrink as a site grows, so the multiple is the easiest number to read and to compare between crawls.
How does the simulator treat nofollow links?
By default a nofollow link is not followed, but it still counts in the source page's denominator, so its share of equity evaporates instead of being redistributed to the other links. That matches what Google said in 2009 when it confirmed that PageRank sculpting with nofollow no longer works. You can switch to removing nofollow links entirely or treating them as followed to compare the outcomes.
Which damping factor should I use?
Leave it at 0.85 unless you have a reason to change it. That is the value most often quoted from the original PageRank work and the default in most graph libraries. A lower value spreads scores more evenly; a higher value rewards deep link structures and takes longer to converge. The more useful test is whether your conclusions hold at 0.75 and at 0.90. If a page is weak at every setting, the finding is robust.
Does the simulator include backlinks from other websites?
No. It models internal links only, so every page starts with the same baseline and equity only moves between your own pages. In reality, pages with strong external links bring more authority into the site than the model shows. You can approximate this by adding your most-linked pages as seeds, which concentrates the starting equity on them.
How large a site can it handle?
The target is one million link rows and around 100,000 pages on a typical laptop, with parsing and calculation finishing in well under a minute. Larger exports may still work, but they depend on how much memory your browser can use. Filtering the export to hyperlinks only before uploading makes a noticeable difference.