How to Use Frase AI for Internal Linking Plans: Ready-to-Use Templates

Tariq edits a 140-post blog about balcony and small-garden growing. Traffic has plateaued, several good guides get almost no internal links, and a new "grow herbs indoors" cluster is about to launch. He has an export of URLs and titles from his CMS and a Friday deadline for a linking plan his two freelance writers can follow. This guide shows how to build Frase AI internal linking plans that are specific and checkable, using filled-in templates you can adapt.

Where Frase fits, and the line you must not cross

Frase is a content research and optimisation tool. It is generally used for building briefs from search results, analysing topics, drafting and scoring content, and, depending on your plan and its current features, suggesting internal links. Check the tool's current features, limits and plan details before relying on any specific function.

What an AI assistant does well here: grouping pages into clusters, proposing anchor-text variations, spotting pages that should logically link to each other, and turning a spreadsheet into an action list.

What it cannot do: know your real site structure unless you supply it, see which pages actually get traffic or conversions, or guarantee ranking effects. The biggest risk is invented URLs: if you do not give a model a list of pages, it will make up plausible ones. Every prompt below supplies the URL list and forbids anything outside it.

Internal links are a structural and editorial decision. Links should help readers move to a genuinely related page; do not stuff exact-match anchors or create links for their own sake.

The internal linking prompt formula

1. Inventory: paste a table of real URLs, titles, and one-line topics (from your CMS export or crawl).
2. Goal: support a new cluster, rescue orphan pages, or consolidate overlapping posts.
3. Rules: only use listed URLs; maximum links per page; where in the page links go.
4. Anchor policy: descriptive, varied, no exact-match repetition.
5. Output: a table with source URL, target URL, anchor, placement and reason.
6. Verification: ask it to flag low-confidence matches.

Prepare your inventory first. A simple pasted table works:

URL | Title | Primary topic | Word count | Last updated
/herbs-indoors-beginners/ | How to Grow Herbs Indoors for Beginners | indoor herbs | 2100 | 2025-03
/basil-pot-size/ | What Size Pot Does Basil Need? | basil containers | 900 | 2024-08
/grow-lights-windowsill/ | Do You Need a Grow Light on a Windowsill? | grow lights | 1500 | 2025-01

Every example below assumes you have pasted an inventory like this. For the full set of tools around content workflow, AI prompts for workflow automation shows how to turn repetitive steps like this into a routine.

Audit and orphan-page prompts

Find orphan and under-linked pages

You are an SEO editor. Below is my site inventory (URL | Title | Topic | Word count | Last updated) for a 140-post balcony and small-garden blog, followed by a second list of internal link counts per URL exported from my crawler (URL | inbound internal links).

Task: 1) List every page with 0 or 1 inbound internal links. 2) For each, say which topic cluster it belongs to based on the listed title and topic only. 3) Rank them by how valuable they probably are for readers (guides first, thin posts last) and give a one-line reason. Do not invent URLs, titles or link counts. If you cannot decide, write "unclear".

Inventory:
[PASTE]
Link counts:
[PASTE]

Why it works: It uses real link counts and bans inventions, so the audit rests on your data. What to expect / check: A ranked list. Spot-check five rows against the crawler. Iterate: "For the top ten, propose two candidate source pages each from the inventory and explain the topical fit."

Cluster mapping

Group the URLs in this inventory into topic clusters. For each cluster give: a short name, the proposed pillar page (the broadest, most useful guide in the list), the supporting pages, and any pages that do not fit. Use only listed URLs. Do not rename pages. Where two pages seem to compete for the same topic, flag them as possible overlap rather than choosing a winner.

Inventory:
[PASTE]

Why it works: Flagging overlap keeps it from making silent consolidation decisions. What to expect / check: Clusters like "herbs", "containers", "lighting". Check that the pillar is genuinely the broadest page. Iterate: "Show the clusters as a table with columns: cluster, pillar URL, number of supporting pages, overlap flags."

Plan-building prompts

Linking plan for a new cluster launch

We are launching a new cluster about growing herbs indoors. The pillar page will be /herbs-indoors-beginners/ (published). Supporting pages are listed below, some existing and some planned (marked PLANNED). Build an internal linking plan using only these URLs.

Rules: each supporting page links to the pillar once, near the first 150 words, with descriptive anchor text; the pillar links to each supporting page once in a relevant section; supporting pages link to at most two sibling pages where a reader would naturally need them; no more than 8 internal links on any page; no repeated identical anchor text across the plan. Output a table: source URL, target URL, suggested anchor, placement note, reason, confidence (high/medium/low).

URLs:
[PASTE LIST WITH PLANNED MARKED]

Why it works: Quantified rules (once, first 150 words, max 8) make the plan reviewable and consistent for writers. What to expect / check: A table with 30 to 60 rows. Open the pages and read sentences around suggested placement; confirm each link makes sense to a reader. Iterate: "Remove every row with low confidence and list what extra information would raise it."

Link from old high-traffic posts to new content

These five existing posts get steady traffic (URLs and titles below). Suggest where each could link to a new page. For each existing post, propose at most two links to pages from the new-content list, name the section of the old post where the link fits based on its title and topic, and write the sentence you would add (max 25 words, natural, no promotional language). Use only listed URLs. If the topic fit is weak, say "no good fit".

Existing posts:
[PASTE]
New pages:
[PASTE]

Why it works: Asking for the actual sentence turns the plan into an editor's task. What to expect / check: You must open each old post to confirm the placement exists; the model only sees titles. Iterate: "Give a second sentence option per link with a different anchor."

Anchor text variations

For the target page /grow-lights-windowsill/ (title: "Do You Need a Grow Light on a Windowsill?"), propose 8 anchor text options. Mix: 2 partial-match, 2 descriptive phrases, 2 question-style, 2 branded or generic-contextual (for example "our windowsill lighting guide"). Each under 6 words. Do not repeat the exact title. Mark which are best for in-paragraph use and which for lists. Avoid "click here" and "read more".

Why it works: Variation guidance protects against repetitive exact-match anchors. What to expect / check: Read each anchor in a sentence; discard awkward ones. Iterate: "Give three example sentences from a basil care article that use the best anchors naturally."

Placement and writing prompts

Insert links into a draft

Below is a draft article and a list of approved internal links (URL and preferred anchor). Insert the links into the text where they fit naturally by rewriting at most one sentence per link. Do not add any link that is not on the list. Do not change facts. Show the revised paragraphs only, with the new link in markdown, and list the changed sentences separately so I can review them.

Draft:
[PASTE]
Approved links:
[PASTE]

Why it works: Whitelisting links and showing changed sentences keeps the editor in control. What to expect / check: Verify no invented URL appears and that tone remains unchanged. Iterate: "Remove any link that interrupts a step-by-step instruction and propose a 'related reading' line instead."

Navigation and hub page content

Write an introduction and annotated link list for a hub page called "Indoor herb growing hub". Use only these URLs and titles: [PASTE]. For each link write a one-sentence description (under 20 words) telling the reader when this page helps. Group links under three headings that make sense for the listed topics. Do not claim any page contains information not suggested by its title.

Why it works: Annotating links by reader need makes the hub useful, not just a list. What to expect / check: Check each description against the page itself. Iterate: "Reorder groups from beginner to advanced and rewrite headings accordingly."

Cleaning up overlap and cannibalisation

Two posts may compete: "What Size Pot Does Basil Need?" (/basil-pot-size/) and "Best Containers for Growing Herbs" (/best-herb-containers/). Based only on titles and the summaries below, recommend one of: keep both and cross-link, merge into one, or differentiate with sharper focus. For each option give the reader-facing reason, the risk, and the redirect implication. Do not assume any ranking or traffic data; list what data I should check first in my analytics or search console.

Summaries:
[PASTE]

Why it works: It makes the model reason about options and the evidence required rather than pretending to know performance. What to expect / check: A decision framework. Check real data before merging or redirecting. Iterate: "Draft the redirect map and the update checklist for the merge option."

Worked example: weak to strong

Prompt 1 (weak):

Give me an internal linking plan for my gardening blog.

Typical result: Generic advice (link to related posts, use keyword anchors) plus a table of made-up URLs like /best-herbs-for-balcony/ that do not exist on the site. Following it would produce broken links.

Prompt 2: The cluster launch prompt with the real inventory and rules.

Why it is better: It is anchored to real URLs, quantified rules and an output table with a reason and confidence for every link, so a writer can implement or reject each one.

End-to-end workflow

1. Export URLs, titles and inbound link counts from your CMS and crawler.
2. Run the orphan audit prompt.
3. Run cluster mapping and agree pillars with your team.
4. Build the plan for the highest-value cluster using the launch prompt.
5. Use the anchor and insertion prompts as writers edit.
6. Re-crawl after publishing and check for broken links and orphan changes.

If your site is a service business site rather than a blog, Hostinger AI prompts for local service business websites covers page structure that makes linking simpler.

Troubleshooting

| Problem | Likely cause | Fix | |—|—|—| | Output includes URLs not on your site | No inventory supplied | Paste the list and add "use only listed URLs" | | Links feel forced | Plan driven by keywords not reader needs | Require a one-line reader reason per link | | Too many links per page | No cap | Set a maximum and ask for ranking by relevance | | Repetitive anchors | No variation rule | Add an anchor policy and a no-repeat rule | | Placement impossible | Model only saw titles | Have an editor confirm placement in the page | | Plan too large to implement | No prioritisation | Ask for a top-20 list by reader value |

Review checklist

• Does every target URL exist and return a normal page (not a redirect or error)?
• Does each link help a reader, based on the surrounding sentence?
• Are anchors descriptive and varied?
• Are link counts per page reasonable, with no clusters of links in one paragraph?
• Did you avoid linking to thin, outdated or noindexed pages?
• Is the plan tracked so you can measure changes with your analytics or search console?

FAQ

Does Frase automatically add internal links?

Check its current features and your plan. Whatever it offers, review suggestions against your real site before publishing.

How many internal links per page are right?

There is no universal number. Use as many as help readers, and avoid clusters of unrelated links.

Will internal links improve rankings?

They help search engines and readers discover and understand pages, but outcomes depend on many factors. Do not expect guaranteed results.

Can I paste my whole sitemap?

For large sites, split it by section so the model can handle it. Check the tool's input limits.

Should I use exact-match anchors?

Use natural, descriptive anchors and vary them. Avoid mechanical repetition.

Conclusion

A good internal linking plan is built from your real URL inventory, clear rules, and a reader reason for each link. Use Frase and AI assistants to speed up grouping, anchors and drafting, but verify every URL and placement yourself before publishing.