Internal linking is the cheapest ranking boost I know of. No outreach, no budget, no waiting on anyone else. You already own both ends of every link. And yet on most sites I audit, it’s the most neglected part of SEO, because doing it properly by hand across 90 articles is miserable work.
So I hand it to Claude. Once a week, Claude reads my Google Search Console data, works out which phrases each article already ranks for, finds places in my other posts where those phrases come up naturally, and turns them into links. I approve the plan, Claude makes the edits on the live site, and I get a changelog. I run this on my own sites and on my clients’ sites, and below is the exact process and prompt.
Why Search Console data makes the links better
Most internal linking advice tells you to link “related” articles. Related according to whom? You end up guessing at anchor text, and you usually guess the keyword you wanted the page to rank for, not the one Google actually ranks it for.
Search Console removes the guessing. If my post on scheduling Instagram posts sits at position 6 for “schedule instagram reels,” then that phrase is the anchor text Google already associates with that page. Linking to it with that phrase from other relevant posts reinforces something Google already believes, which is a lot easier than convincing it of something new.
It also matches Google’s own guidance. Their link best practices say good anchor text should be descriptive, reasonably concise, and relevant to both pages, and they warn that if it feels like you’re forcing keywords in, it’s probably too much. A query a page genuinely ranks for, used where it already appears in your text, passes that test almost by default.
Get your Search Console data to Claude
You have two options. The simple one is an export: in Search Console, open Performance, set the date range to the last three months, make sure both Queries and Pages are included, and export. Three months is enough data to be reliable without being dominated by rankings you lost long ago.
The better option is an MCP connector, so Claude can pull the data itself every week without you exporting anything. There are several open-source Google Search Console MCP servers on GitHub that work with Claude Desktop and Claude Code. If you already pay for Ahrefs, its MCP connector also exposes your GSC keywords and pages, which is what I use. Either way, what Claude needs is the query, the page, the position and the impressions.
Give Claude access to your website
This is the step that turns a nice spreadsheet into actual changes. My sites run on WordPress, and I use a custom MCP plugin that lets Claude list posts, read their content and edit them. Without it, Claude can only tell you where the links should go, and you’re back to doing the editing yourself.
WordPress isn’t the only option. There are MCP connectors for most blogging and CMS platforms now, including Webflow, Ghost, Sanity and Contentful, and if your blog is a folder of Markdown files in a Git repo, Claude Code can edit those directly with no connector at all.
One warning. An MCP with write access can change your live site, so the prompt below forces Claude to show you the plan and wait before touching anything. Don’t remove that step. And make sure your CMS keeps revisions (WordPress does by default) so any edit can be rolled back.
The prompt I use
Copy this, swap in your site name, and attach the export or point it at your GSC connector:
I've attached my Google Search Console export (or: pull it from the GSC connector) for the last 3 months. My website is connected through the [site name] MCP. 1. For each article, find the queries where it ranks best (position 1–10, decent impressions). Those are its target anchors. 2. Go through every published post. Wherever the text naturally mentions another article's top query (or a close variant), turn that phrase into a link to that article. 3. Rules: - Max 3–5 new internal links per post, spread through the body (no headings, intro sentence, or existing links) - Never link a post to itself, and only link each target once per post - Anchor text must read naturally. Don't rewrite sentences just to force a keyword in - Prioritize sending links to pages ranking 4–15 (they gain the most) 4. Before changing anything, show me a table: source post → anchor text → target URL → target's query and position. Wait for my OK. 5. After I approve, make the edits through the MCP, then give me a changelog so I can revert if needed.
A few of those rules are there because I learned the hard way.
The cap of three to five links per post stops Claude from turning every article into a wall of blue text. Without it, a 2,000-word post can easily pick up fifteen new links in one pass, which is bad for readers and dilutes every link on the page.
Prioritizing pages in positions 4 to 15 is the part that actually moves rankings. A page already at position 1 or 2 doesn’t need help. A page stuck at position 8 or 12 is close enough that a few relevant internal links can push it onto page one or into the top three, where the clicks are.
“Don’t rewrite sentences to force a keyword in” is the rule I’d keep if I could only keep one. Left to its own devices, an AI will happily rewrite your sentence to fit the anchor, and you end up with prose that reads like 2012 SEO. I only want links where the phrase is already there.
And the approval table is non-negotiable. Read it. Most rows will be fine, but every week I reject a couple, usually where a phrase technically matches but the target page doesn’t answer what the reader wants at that point in the article.
Then leave the site alone
After the edits go live, don’t touch anything for a few days. Google needs to recrawl the changed pages before anything shows up in rankings, and if you keep changing things you’ll never know what caused what.
Then repeat it every week. Each run uses fresh Search Console data, so as pages move up, new pages enter the 4 to 15 range and get their turn. New articles you published that week get linked from older ones too, which helps them get crawled and indexed faster. Over a couple of months you’ll see pages climb, and the rankings report starts to look very different.
You can do all of this by hand
Claude just makes it fast. The manual version is the same logic: export Search Console, sort by page, write down each article’s best-ranking queries in positions 4 to 15, then search your own site for those phrases (a Google search like site:yourdomain.com "schedule instagram reels" works) and add links where the phrase already appears. It takes a few hours per round on a 50-post site instead of ten minutes, but the results are the same.
Internal links won’t carry you alone
Internal linking redistributes the authority your site already has. It can’t create more. If your whole domain has very few backlinks, you’re shuffling around a small pile, so keep doing backlink outreach alongside it. And keep looking for new content opportunities: the same Search Console export usually shows queries you get impressions for but have no dedicated article about. Those are your next posts, and once they’re published, next week’s internal linking run will start sending links to them.
Internal linking FAQs
How many internal links should a blog post have?
There’s no fixed number, but I add three to five new contextual links per post per round, and I never link the same target twice in one post. What matters more than the count is that every link sits in a sentence where a reader would actually want to click it.
Should internal link anchor text be an exact-match keyword?
For internal links, using the query a page already ranks for is fine and helpful, as long as it reads naturally in the sentence. Google recommends descriptive, concise anchor text and warns against stuffing keywords. Close variants are good too, and they keep your anchors from looking robotic.
Which pages should get the most internal links?
Pages ranking roughly between positions 4 and 15. They’re close to the top results, so a small boost produces a visible gain. Pages already in the top two don’t need much help, and pages beyond position 30 usually have a content problem that links won’t fix.
Is it safe to let AI edit my website?
It is if you keep a human approval step and your CMS keeps revisions. My prompt makes Claude show every proposed link in a table and wait for my OK, then produce a changelog afterward. I’d never run it in a fully automatic mode on a client site.
How long does it take for internal links to affect rankings?
Google has to recrawl the changed pages first, which can take a few days to a couple of weeks depending on how often your site gets crawled. Give each round at least a week before judging it, and look at the trend over a month or two, not day-to-day movement.
Do I need a paid tool for this?
No. Google Search Console is free, the export is free, and you can run the prompt in any Claude plan that supports connectors. A paid SEO tool like Ahrefs is a convenience, not a requirement.
Does internal linking help with GEO and AI search?
Indirectly, yes. AI search tools lean on pages that rank well in traditional search, so anything that lifts your rankings raises your odds of being cited. Clear internal links also make it easier for crawlers, including AI crawlers, to find and understand all your pages.