You don’t need to become a prompt engineer to get useful SEO work out of ChatGPT, you just need the right prompts already written for you. This article gives you fifteen of them, grouped by the task you’re actually doing, so you can copy, paste, swap in your details, and move on.
- One copy-paste prompt gets you SEO titles plus a matching meta description in a single shot.
- Good SEO prompts get grouped by task, research, outlines, on-page, meta, FAQ, and internal links, instead of one generic “write me SEO content” ask.
- Getting cited in AI Overviews and ChatGPT answers needs its own prompt style: a self-contained 40-60 word answer under every H2, not just a Google-ranking prompt.
- ChatGPT can’t be trusted for real search volume or keyword difficulty numbers, those still come from a dedicated SEO tool.
The 3 ChatGPT Prompts to Copy First
These three show up in almost every SEO workflow: titles and metas, a fast outline, and an FAQ block. Grab whichever one matches what’s on your plate right now.
Prompt: Act as an SEO copywriter. Write 5 title tag options (60-70 characters each) and one matching meta description (140-160 characters) for a page about [topic], targeting the keyword [keyword]. Present the results in a table with a character count column.
When to use it: right before publishing, or when refreshing an old page’s metadata.
Variables: [topic], [keyword], target audience (optional).
Prompt: Act as a content strategist. Create a blog outline for the keyword [keyword], targeting [audience], around [word count] words, with the goal of [page goal]. Include an H1, H2s, and one-line notes on what each section should cover.
When to use it: at the start of a new article, before you open a blank doc.
Variables: [keyword], [audience], [word count], [page goal].
Prompt: Act as an SEO content writer. Write 5-7 FAQ questions and 2-3 sentence answers for a page about [topic], in a [brand tone] voice, suitable for FAQPage schema and People Also Ask.
When to use it: when closing out an article or building a standalone FAQ page.
Variables: [topic], [brand tone], target keyword (optional).
Keyword and Topic Research Prompts
ChatGPT is decent at brainstorming keyword angles and terrible at telling you which ones actually get searched. Use these prompts to:
ChatGPT is decent at brainstorming keyword angles and terrible at telling you which ones actually get searched. Use these prompts to expand your topic universe with keyword variations and related terms, group ideas by search intent to spot content gaps, and verify actual search volume and difficulty numbers elsewhere in a dedicated SEO tool.
Prompt: Act as an SEO researcher. Generate 20 long-tail keyword variations for the seed keyword [seed keyword], grouped into 3-5 topic clusters by search intent (informational, commercial, transactional).
When to use it: early planning, when you have a broad topic but no content calendar yet.
Variables: [seed keyword], industry or niche context.
Prompt: What related terms, entities, and subtopics should a thorough page about [topic] cover to be seen as comprehensive by both readers and search engines?
When to use it: instead of the outdated “give me LSI keywords” ask, which chases a concept Google has confirmed it doesn’t actually use for ranking, according to Search Engine Land’s prompt guide.
Variables: [topic], competitor URLs (optional).
One honest limit here: ChatGPT can only guess at monthly search volume and keyword difficulty. Those numbers still need to come from a real SEO tool. ChatGPT is built for expanding the idea list, not scoring it.
Content Brief and Outline Prompts
The outline prompt above works fine for a quick draft, but a proper brief needs more scaffolding, especially if you want the output to sound like your brand instead of a generic AI voice. Before running the prompt below, paste in one or two examples of your best-performing content and a line about your brand voice. That single step is what separates a usable brief from a bland one.
Prompt: Act as a senior content strategist. Build a full content brief for the keyword [keyword]. Include: related terms and entities to cover, target word count of [word count], target audience of [audience], the primary goal of the page ([page goal]), a suggested H1/H2 structure, and 3 competitor angles this piece should differentiate from. Reference these examples of our existing content for tone: [paste 1-2 examples].
When to use it: for any article that will get assigned to a writer or go straight into a CMS, since it front-loads the context a bare outline skips.
Variables: [keyword], [word count], [audience], [page goal], brand voice examples.
This mirrors guidance from content teams who write these prompts daily: give the model clear instructions, feed it reference material, and break a complex task like “write my brief” into the specific sub-parts you actually need.
On-Page, Title and Meta Description Prompts
Two different jobs live under this heading:
One prompt reviews content that already exists for gaps and issues, the other generates fresh titles and metas for new pages, and you should keep them separate so the output doesn’t blur into a vague “make this better” response.
- One prompt reviews content that already exists for gaps and issues.
- The other generates fresh titles and metas for new pages.
- Keep them separate so the output doesn’t blur into a vague “make this better” response.
Prompt: Act as an SEO auditor. Review this content: [paste content]. Identify missing subtopics, weak or missing internal linking opportunities, heading structure issues, and keyword usage gaps for the target keyword [keyword].
When to use it: on a published page that’s underperforming and needs a refresh pass.
Variables: [content], [keyword].
Prompt: Generate 5 title tag and meta description pairs for a page about [topic], targeting [keyword]. Titles must be 60-70 characters, meta descriptions 140-160 characters. Present as a table with a character count column.
When to use it: for new pages, or whenever you need multiple options to A/B test.
Variables: [topic], [keyword].
ChatGPT counts characters unreliably on longer strings. That’s exactly why both prompts above ask for a table with a character count column: it’s the fastest way to see what’s actually going to get cut off in a search result.
FAQ Blocks and Internal Linking Prompts
These two tasks get lumped together a lot, but they solve different problems and deserve their own prompts: one builds reader-facing FAQ content, the other builds the internal link map underneath it, and for FAQ content you should feed ChatGPT your brand tone directly rather than describing it abstractly since it produces a noticeably tighter result.
- One builds reader-facing FAQ content.
- The other builds the internal link map underneath it.
- For FAQ content, feed ChatGPT your brand tone directly rather than describing it abstractly.
- It produces a noticeably tighter result.
Prompt: Act as an SEO content writer. Write 5-7 natural FAQ questions and answers about [topic], in the voice shown in this example: [paste 1-2 examples of existing content]. Keep answers to 2-3 sentences, written for both readers and FAQPage schema.
When to use it: whenever the outline calls for an FAQ section, or you’re building a standalone FAQ page.
Variables: [topic], brand voice sample, target keyword (optional).
Internal linking is a separate ask, and it works best when you hand ChatGPT the actual list of URLs plus the source content, so it can match anchor text to real pages instead of inventing links.
Prompt: Act as an SEO specialist. Here is a list of URLs on our site: [paste URL list]. Here is the content of a new article: [paste content]. Suggest 3-5 internal links from the URL list that fit naturally into this article, with keyword-optimized anchor text for each.
When to use it: right before publishing, once the draft is close to final.
Variables: [URL list], [content].
GEO Prompts: Writing to Get Quoted by AI Overviews and ChatGPT
AI Overviews now show up on roughly half of tracked Google searches, and ChatGPT itself passed 800 million weekly users in October 2025. That makes ChatGPT a search surface in its own right. Ranking on page one doesn’t guarantee an AI answer will quote you, so writing to actually get quoted needs its own prompts.
Good to know: Generative engine optimization, or GEO, is the practice of structuring content so AI answer engines can lift it directly into a response. It sits alongside traditional SEO: the ranking still matters, and the paragraph also has to be quotable on its own.
Prompt: For each H2 in this outline: [paste outline], write a self-contained 40-60 word answer that fully answers the heading’s question on its own, without needing the surrounding paragraphs for context. Write it so an AI search engine could quote it directly.
When to use it: right after you’ve drafted your headings, before writing the full body copy under each one.
Variables: outline or list of H2s, topic.
Prompt: List the 5 most common follow-up questions a reader would ask after getting an initial answer about [topic]. Answer each in 2 sentences, written to stand alone if quoted by an AI assistant.
When to use it: when building FAQ sections or a “people also ask” style block aimed at AI citation, not just search snippets.
Variables: [topic], primary keyword.
Once the content is live, checking whether it’s actually getting pulled into AI answers is its own separate step, worth reading up on through resources like this guide to getting AI models to cite your content. Some teams run these prompts one at a time by hand.
Others hand the whole loop to an autonomous agent: research, drafting with this GEO structure baked in, and publishing. Claire is one option built around that workflow, alongside plenty of teams who prefer to keep running the prompts themselves.
Common Prompting Mistakes That Produce Vague SEO Output
A handful of habits explain nearly every flat, generic ChatGPT draft, and all of them are easy to fix once you spot the pattern.
- Vague instructions produce generic content. Paste in reference material every time.
- Guessing search volume or keyword difficulty produces often-wrong numbers, so keep those metrics in a dedicated SEO tool.
- Asking for “LSI keywords” chases a concept Google doesn’t use; use the related-terms-and-entities prompt instead.
A simple tagged doc, a board, or a ChatGPT Projects folder organized by task, research, on-page, GEO, works fine for that last point. If you want a ready-made structure for it instead of building one from scratch, this prompt library template is built for exactly that.
Where the Human Still Needs to Step In
Every prompt in this article gets you a strong first draft, not a finished, publishable page. The gap between the two is the part ChatGPT can’t close on its own: checking that the facts and numbers are actually correct and making sure the tone still sounds like your brand. That review pass isn’t optional, and skipping it is how AI-generated pages end up flagged for being generic or wrong.
The practical next step is deciding how much of this loop you want to keep running by hand. If you’re saving these prompts and running them yourself, set up that tagged library now so you’re not hunting for them next week. If you’d rather someone (or something) else handled research through publishing end to end, that’s the gap tools built around GEO-aware writing are designed to close, Claire is one honest option among them, alongside plenty of teams who keep the human review pass and just run the prompts faster.
FAQ
Can I use these same prompts in Claude or Gemini, or are they ChatGPT-specific?
Yes, every prompt here works in Claude, Gemini, or any other capable chatbot with little to no editing. The structure, role, task, variables, format, is what matters, not the specific tool running it. You may need to nudge Gemini or Claude slightly differently on table formatting, but the core instructions transfer directly.
How many of these prompts should I run before I have a full article ready?
For a standard article, expect to use three to five: a keyword or topic research prompt, the outline or brief prompt, one GEO answer prompt per section, and the title/meta prompt at the end. FAQ and internal linking prompts get added on top if your outline calls for those sections.
Do I need ChatGPT Plus/Team, or does the free version work for these prompts?
The free tier handles all fifteen prompts fine, since none of them depend on paid-only features like advanced data analysis or larger file uploads. A paid plan mainly helps if you’re pasting in long reference documents or running these prompts at high volume throughout the day.
Should I write one master prompt that does everything at once instead of running several separate ones?
No, a single do-everything prompt tends to produce shallower results on every part of the task. ChatGPT handles a focused ask, like just the outline, or just the FAQ block, more thoroughly than a request stacked with five jobs at once. Running the prompts in sequence, checking each output before moving to the next, gets you a better finished piece than one giant combined prompt.
Can I chain these prompts together in one ChatGPT conversation, or should each run separately?
Chaining works well within one topic: run the outline prompt, then feed that outline straight into the GEO answer prompt and the FAQ prompt in the same thread, since ChatGPT keeps the context. Switch to a fresh conversation once you move to a different article or keyword, otherwise earlier context tends to bleed into outputs that should stay focused on the new topic.