Copy-Paste AI Prompts for SEO Meta Titles, Meta Descriptions and LinkedIn Posts

Most teams discover generative AI through the wrong task: they ask it to write whole articles, get generic prose, and conclude the tools are overrated. The higher-leverage entry point is the small, structured copy you produce constantly — meta titles and descriptions that decide whether a searcher clicks your result, and LinkedIn posts that decide whether anyone sees the content you worked hard on. Both have tight length limits, clear success criteria and a repeatable format, which makes them near-perfect prompt targets.

The catch is that a lazy prompt ("write me a meta description for my SEO page") produces lazy output. The prompts below all follow the same five-part skeleton, and it is the skeleton — not the tool — that does the work:

• Role — tell the model what it is ("you are an SEO copywriter"), which sets vocabulary and priorities.
• Context — the page, audience, keyword and differentiator. The model cannot know your business; feed it.
• Constraints — character limits, banned phrases, tone rules. Constraints are what separate usable output from slop.
• Format — ask for a table or labelled options so the output drops straight into your workflow.
• Guardrail — an explicit instruction not to invent facts, and to ask when information is missing.

Copy the templates as-is, replace the [BRACKETED] placeholders, and edit the rules to match your house style. Every one of them is drilled hands-on in the WSQ Generative AI for SEO course, where a trainer corrects your prompts on your own site's pages.

Part 1 — Prompts for SEO meta titles and descriptions

Prompt 1: Meta title and description for a single page

The workhorse. Asking for three options in a table forces variety and makes the character counts checkable at a glance.

You are an SEO copywriter. Write 3 meta title options (max 60
characters each) and 3 meta description options (max 155 characters
each) for the page below.

Page topic: [WHAT THE PAGE IS ABOUT]
Primary keyword: [KEYWORD] - it must appear near the start of each title
Audience: [WHO SEARCHES FOR THIS]
Differentiator: [WHY THIS PAGE - e.g. price, funding, location]

Rules:
- No clickbait and no claims the page cannot back up
- Include one concrete number or specific detail in each description
- End each description with a soft call to action
- Return a table: option, title, title characters, description,
  description characters

What to check: the model counts characters unreliably — paste the winner into a pixel-width checker before shipping, and confirm every claim ("from $292.50", "2-day course") is actually true on the page.

Prompt 2: Bulk metas for a whole section of the site

Where AI genuinely changes the economics: a hundred course or product pages that would take a week by hand take an afternoon with review.

You are an SEO assistant. I will paste a table of URLs with columns:
URL, page type, primary keyword, current meta title.

For each row write a new meta title (max 60 chars) and meta
description (max 155 chars).

Rules:
- Keep each row's primary keyword in the first half of the title
- Never reuse the same opener twice - vary the phrasing across rows
- Append " | [BRAND]" only where it still fits the limit
- Flag any two rows that target the same query (cannibalisation risk)
- Return one table I can paste into a spreadsheet

[PASTE YOUR TABLE]

What to check: spot-check ten rows, not two. Bulk runs fail in patterns — if the model drifts into a repeated formula halfway down, regenerate from that row with "vary the openers" restated.

Prompt 3: Rescuing a page with high impressions but low CTR

Pull the numbers from Google Search Console first — this prompt turns a reporting insight into a rewrite, which is the single fastest SEO win available to most sites.

Here is Google Search Console data for one page (last 3 months):

Query: [TOP QUERY]   Impressions: [N]   Clicks: [N]
CTR: [N]%   Average position: [N]

Current meta title: [TITLE]
Current meta description: [DESCRIPTION]

The page ranks well but searchers skip it. Diagnose the likely reason
in 2 sentences, then write 3 alternative title/description pairs for
the same query: one leading with a number, one leading with the
outcome, one leading with a question. Max 60 / 155 characters.

What to check: the diagnosis is a hypothesis, not a fact. Ship one variant, wait three weeks, and compare CTR in Search Console before rolling the pattern out.

Prompt 4: Local-intent metas for Singapore searches

Singapore searchers respond to funding signals the way other markets respond to discounts — but only claim them where they are true.

Write meta titles and descriptions for this page targeting searchers
in Singapore.

Page: [PAGE TOPIC]
Primary keyword: [KEYWORD] singapore

Rules:
- Mention Singapore naturally in the title, not bolted on
- Include one trust or funding signal in the description, e.g.
  "SkillsFuture Credit eligible" or "up to 70% WSQ funding",
  ONLY if true for this page
- Max 60 / 155 characters
- Give 3 pairs ranked by expected click-through rate, with one line
  explaining your ranking

What to check: funding claims are regulated territory — verify the exact scheme names and percentages against the live page before publishing. A wrong subsidy figure in a meta description is a complaint waiting to happen.

Part 2 — Prompts for LinkedIn posts

Prompt 5: Turn an article into a LinkedIn post

The most common repurposing job. The two rules that matter most: the hook must survive the "…see more" fold, and the post must contain one concrete number so it reads like experience rather than content marketing.

Turn the article below into a LinkedIn post for my company page.

Structure:
- Hook: the first 2 lines must work before the "...see more" fold -
  no throat-clearing
- Body: 3-5 short paragraphs, one idea each, plain English
- Include one concrete example or number from the article
- CTA: one line inviting readers to the full article at [URL]
- 3-5 hashtags: one broad, the rest niche

Rules: no emojis in the first line, never open with "I'm excited to
share", 150-220 words total.

[PASTE ARTICLE OR KEY POINTS]

Prompt 6: A first-person post from something you learned

Personal posts outperform company posts on LinkedIn, and the guardrail line here is the difference between authentic and invented.

Write a first-person LinkedIn post based on this experience:

What happened: [1-3 SENTENCES ON WHAT YOU DID OR LEARNED]
Who it helps: [AUDIENCE]
The takeaway: [THE ONE THING READERS SHOULD REMEMBER]

Rules:
- Open with the moment of surprise or failure, not the lesson
- Short lines, generous white space, max 3 hashtags
- Do NOT invent details - if a fact is missing, ask me instead
- End with a genuine question that invites comments, not "thoughts?"
- 120-180 words

Prompt 7: Announcing a course run, webinar or event

Announcement posts die when they lead with the thing instead of the problem. This template forces the order right and keeps the funding concrete.

Write a LinkedIn post announcing this course run for a company page.

Course: [COURSE NAME]
Date and format: [DATE, CLASSROOM / ONLINE]
Who it is for: [AUDIENCE]
Funding: [E.G. UP TO 70% WSQ FUNDING, SKILLSFUTURE CREDIT ELIGIBLE]
Registration link: [URL]

Rules:
- Lead with the problem the course solves, not the course name
- One line on what participants build or walk away with
- State the funding in numbers - figures beat adjectives
- Single CTA line with the link, under 150 words, max 4 hashtags

Prompt 8: One article, five posts — a week of content

Repurposing is where consistency comes from. Asking the model to vary the format across the week stops your feed reading like a template.

Take the article below and plan 5 LinkedIn posts from it, one per
weekday.

For each post give: the angle, the opening hook line, and 2-3 bullet
points of body content. Vary the format across the week: one how-to,
one contrarian take, one mini case study, one list post, one question
post. Never repeat the same hook structure twice.

Flag which single post you would put paid spend behind, and why.

[PASTE ARTICLE]

The pre-publish checklist

Whatever the prompt produced, it goes through the same 60-second review before it ships:

Learn the full workflow, with WSQ funding

Templates get you started; the skill is knowing how to adapt them to your pages, your audience and your data — and how to chain them into a complete SEO workflow from keyword research to content production to measurement. That is exactly what WSQ Generative AI for Search Engine Optimization (SEO) covers: a 2-day, 16-hour beginner-level course spanning SEO and generative AI fundamentals, AI-powered keyword research and content strategy, creating SEO-optimised content with generative AI, AI for on-page optimisation, and SEO performance, ethics and best practices.

Because it is a WSQ course, the funding is substantial for eligible learners:

• Full fee $750 before GST.
• $442.50 nett for Singapore Citizens and PRs aged 21 and above (50% funded).
• $292.50 nett for Singapore Citizens aged 40 and above under MCES, and for eligible SMEs (70% funded).
• SkillsFuture Credit, PSEA and UTAP can offset the fee payable after funding; employers can tap SFEC and claim Absentee Payroll.

Different entry points if this one does not fit: Generative AI for SEO (non-WSQ, shorter, no funding paperwork), WSQ Enhancing Online Presence with AI-Powered SEO, and WSQ Agentic AI for Digital Marketing for the automation layer on top. Browse the full SEO course listing, WSQ Generative AI courses and Content Marketing courses for every option and the next available dates.

Frequently asked questions

View WSQ Generative AI for Search Engine Optimization (SEO) — dates, funding and registration →