A hub, its pages, and the links between them
Every SEO tool draws the topic cluster: a hub in the middle, its pages around it, arrows both ways. SEOLift now reads the arrows off your own pages instead of drawing them, hands the missing ones to the Internal Link Dealer, and checks on the result page whether the topic has room for one more page.
The picture everyone draws and nobody checks
A hub and its pages. For each: does it link to the hub or not. Read, not guessed.
Every SEO tool draws the same picture: a pillar page in the middle, its cluster pages around it, arrows both ways. The picture is right. The trouble is where the arrows come from. In most tools they are a diagram of what you should do, drawn from a keyword list, and nobody checks whether the links exist on your site today.
SEOLift already held every piece of the answer and never joined them. Your Search Console clusters say which page serves which searches. The reading of your site saw every internal link on every page. The cluster map joins the two and asks one question per page: does it link to its hub, and does the hub link back. The answer is read from your pages, not drawn.
A topic is what you already rank for
A topic is the pages you already show for. No keyword list.
Every query cluster in Search Console lands on one page. Pages whose searches open with the same word, or share most of their words, are one topic. The page with the most impressions is the hub and the rest are its supporting pages. No keyword tool and no model decides what belongs together: the grouping is the same rule the briefs use, and it costs nothing.
The home page is left out on purpose. It links to everything and is about everything, so it would be the hub of one blob. Branded searches are left out too. What remains is a handful of topics per site, each with the demand behind it, largest first.
A link in the menu is not a link
For every supporting page the map gives two answers, to the hub and from the hub, and each is one of four words. Linked: the reading saw an internal, followed link in the page. Sitewide: the link is there, but its target is linked from at least 40% of your pages, so it is navigation, not a choice the topic made. Missing: the reading read the page and found no link. Unknown: the reading never reached the page.
Pages that redirect, carry a noindex or point their canonical elsewhere are shown with their state and not asked, because a link to a page Google will not index is not worth placing. The screen names the reading the links were read from and says when a newer one exists. Nothing is fetched and nothing is bought: a shop with tens of thousands of pages is answered from the reading it already has.
Missing links go to the Internal Link Dealer on their own
Missing links go to the Internal Link Dealer on their own. Weighed up, not waved through.
The missing pairs are the list: which page should link to which, ordered by the demand behind the topic, then by the page’s own demand, with the link to the hub first. Without the Internal Link Dealer you download it as CSV and hand it to whoever edits the pages.
With the module, its next run reads the map. A pair from the map is judged first, so the run’s cap can never drop it, and its destination is put in front of the judge even when retrieval would have left it out. Then the usual rules apply: the phrase must already be in a paragraph of the source page, the reader must have a reason, the destination must be a page worth linking to. A cluster pair weighs more, and it can still be refused, with the refusal written down. The run’s report says how many pairs came from the map and how many were placed.
Room for one more page
A topic has a hub and five pages. The result page has room for a sixth. Checked, not guessed.
Under the map there is one more question: does the topic have room for a page you do not have? For the five largest topics SEOLift takes the searches that contain the hub’s own subject from the DataForSEO index, largest first. Then it subtracts: searches you already show for, your brand, your excluded words, searches that leave the hub’s words behind, rephrasings of the hub or of a page you already have, and searches with too little demand or too high a difficulty.
Each survivor is read on the live result page and compared with the hub’s result page and with those of the pages it could be a copy of. Four of ten shared addresses means it is the same search, and it is dropped with the page it duplicates named. What stays becomes a page to write under Pages worth writing, with the hub as the page to link from, and gets a brief overnight. You start it with one button, it says what it will spend first, at most 70 checks, and it stops at your budget.
The seed is always the hub’s own subject. There is no box to type a keyword into, because then the tool would be guessing about demand it has never seen on your site.
What it does not do
It does not ask a model whether two searches mean the same thing. Google already answered by showing the same pages or different ones, and an intersection of two address lists is arithmetic. It draws no links between supporting pages: the evidence that they matter is thinner than for the hub, and they would multiply the list. It writes nothing on your site: the map says which page should link to which, the Link Dealer says what the link should say, and you place it.
And it needs a reading that stored its links. An older reading cannot be mapped, and the screen says to order a new one rather than drawing from memory. All of this sits inside SEOLift, on the Business and Agency plans, under one address: the cluster map.