ChatGPT Prompts for SEO: Briefs, Titles & Schema

ChatGPT prompts for SEO work when you ask for briefs, title options, and schema drafts from facts you already own. This guide gives paste kits for content briefs, title and meta lines, and JSON-LD starters aimed at GPT-5.5 Instant and GPT-5.6 Sol. You leave with ROLE / TASK / FORMAT scaffolds, a SERP_NOTE pack, and claim fences that keep the model from inventing rankings, traffic, or markup Google never asked for.

You still own search strategy and every publish decision. PromptMake /text can scaffold a rough ask at https://promptmake.net/text before you paste into ChatGPT. These kits speed drafting. They do not promise rankings, featured snippets, or traffic.

Who ChatGPT prompts for SEO help

You need a brief that locks intent and outline before a writer drafts, a set of title tags a human can pick from, or a JSON-LD draft you will validate in a rich-results tester. The patterns fit in-house SEO specialists, content leads who hand briefs to freelancers, and marketers who own a small site and write their own pages. You already know the primary keyword, the URL plan, and the pages that should link in. You need the model to turn that pack into a brief, title list, or schema skeleton without inventing Search Console numbers or competitor rankings.

Skip this page if you want a full seven-step article draft formula from intent lock through section writing. That workflow lives in the perfect prompt formula for SEO content guide. Skip it if you want marketing campaign kits or AIDA and PAS copy frameworks. Those libraries live elsewhere on this blog. Skip it if you want the model to invent keyword volume, difficulty scores, or backlink lists you never measured.

Treat ChatGPT as an SEO drafting desk. You supply SERP_NOTE, PAGE_GOAL, and the artifact type. The model proposes brief structure, title variants, or schema fields tied to those facts. You reject any draft that adds rank promises, fake stats, or schema types your page does not support.

Lock SERP_NOTE and claim fences before you prompt

Name the search constraints in the first block of every SEO thread. Editors and engineers punish briefs that invent intent or schema that the live page cannot support. A polished brief from "write an SEO outline for [keyword]" fails the moment a writer ships a comparison page while the SERP is how-to heavy. Put the constraints in ROLE and repeat the hard ones in a REMINDER line so a long keyword paste does not bury them.

Allowed artifacts: content briefs with intent and outline, title tag and meta description option sets, H1 candidates, FAQ question lists tied to People Also Ask notes you paste, and JSON-LD drafts for Article, FAQPage, HowTo, BreadcrumbList, or Organization when your page matches. Banned artifacts: invented keyword volume or difficulty, fake Search Console charts, rank guarantees, "this will hit position one" claims, backlink outreach scripts you did not request, and schema for reviews, products, or events your page does not contain.

Keep a SERP_NOTE sheet outside the chat: primary keyword, secondary terms you own, intent type (informational, commercial, transactional, comparison), top competitor titles and H2 themes you skimmed, questions from People Also Ask you copied by hand, URL slug plan, target word count range, and internal links you can add. Update it when the SERP shifts. Feed it into every prompt. The model should never be your only SERP research tool.

Claim fences belong in every ROLE block. Tell the model: no rank promises, no traffic forecasts, no inventing CTR or impressions, no schema properties without a matching on-page fact, and no stats without a source you paste. Redact API keys, Search Console exports with private URLs you cannot share, and client NDAs before paste. Consumer ChatGPT is a drafting desk, not your analytics warehouse.

Core ChatGPT prompts for briefs, titles, and schema

Strong ChatGPT prompts for SEO give four inputs before any tone request: SERP notes, page goal, the artifact type, and the claim boundary. SERP notes are intent, competitor themes, and questions you copied. Page goal covers audience, CTA, and what the reader must do after reading. The artifact names brief, title pack, or schema draft. The claim boundary forbids rank guarantees, fake metrics, and schema types the page cannot support.

Paste those four blocks near the top. Put format rules next. Put banned outputs at the end so they survive a long paste. GPT-5.5 Instant, GPT-5.6 Sol, Claude Sonnet 5, and Gemini 3.5 Flash all follow labeled blocks well. Vague "optimize this for SEO" prompts produce generic keyword essays because the model has no SERP scan, no URL plan, and no fence against rank theater.

Run one artifact type per thread. Mixing a full content brief and a twelve-variant title pack in the same chat blurs length rules and invents FAQ questions across messages. Close the chat after you finish one pass. Open a new thread for schema so brief marketing language does not leak into JSON-LD field names.

Role, task, format skeleton

Copy this skeleton and fill the brackets with your material:

ROLE: You are an SEO writing assistant for [in-house | agency | indie] teams. You work from SERP_NOTE and PAGE_GOAL only. You write briefs, title options, and schema drafts a human can edit. You never invent keyword volume, difficulty, rankings, traffic, or Search Console numbers. You never promise rankings. You only propose schema types that match PAGE_GOAL and on-page facts I paste.

TASK: Turn SERP_NOTE into a [content brief | title and meta pack | JSON-LD schema draft] for [URL or page type]. Keep every claim tied to SERP_NOTE and PAGE_GOAL. Mark gaps with [NEED FACT]. If I ask for rank guarantees or fake metrics, refuse and offer a brief, title, or schema artifact instead.

FORMAT: Use the section headers listed under OUTPUT_SHAPE. Short sentences. Numbers only when present in SERP_NOTE. Cap each section at the word limits you set.

SERP_NOTE: [primary keyword, secondaries, intent, competitor title themes, PAA questions you copied, slug plan, word count range, internal links you own]

PAGE_GOAL: [audience, reader job-to-be-done, primary CTA, brand voice notes, banned claims]

OUTPUT_SHAPE: [list required sections]

RULES: No rank guarantees. No invented volume or KD. No schema for content the page lacks. If a fact is missing, write [NEED FACT]. Never invent case studies or client results.

REMINDER: Brief, titles, or schema only. Never invent SERP metrics. Never promise rankings.

That reminder line stops the model from padding with "this keyword will rank in 30 days" lines you never earned. Models love tidy growth arcs. Your rule forces a gap marker you can fill from your tools or leave blank.

Paste-ready kits for content briefs

Intent brief pattern adds: "From SERP_NOTE, write a content brief under 400 words. Sections: Intent statement (2 sentences) | Reader job-to-be-done | Must-answer questions (5 from PAA or SERP_NOTE) | Outline (H2 list only) | Differentiation vs competitor themes | Internal links to use | Risks and [NEED FACT] gaps. Ban rank promises and invented volume. Output with those headers only." You paste into Notion, Docs, or your CMS brief field and trim.

Competitive gap pattern: "TASK: From SERP_NOTE competitor themes and PAGE_GOAL, list sections competitors cover that we must match, and sections we add that they lack. Cap each list at 5 bullets. Flag any gap that needs primary research with [NEED FACT]. No traffic claims." Editors use this before they approve the outline.

Snippet target pattern: "TASK: From SERP_NOTE and OUTLINE, label each H2 with a snippet target: paragraph | list | table | FAQ. Propose one 40-60 word direct answer block for the primary query. Ban featured-snippet guarantees. Output as a table: H2 | Snippet target | Notes." Writers draft to the label; you still decide what to ship.

Brief audit pattern: "TASK: Audit BRIEF_DRAFT against SERP_NOTE and PAGE_GOAL. For each claim about intent, competitor gaps, word count, or reader outcome, reply Keep, Soften, or Remove. Soften means the claim overreaches the sources. Flag rank promises. Propose safer wording. Never add new keywords or metrics." Run this before you hand the brief to a writer.

Paste-ready kits for titles and meta

Title pack pattern:

TASK: From SERP_NOTE and PAGE_GOAL, write 10 title tag options under 60 characters each. Primary keyword natural in at least 6. Variants: how-to | list | question | bracket benefit | plain descriptive. Ban clickbait and rank claims. Output as a numbered list with character counts.

Meta description pattern: "TASK: From SERP_NOTE, PAGE_GOAL, and CHOSEN_TITLE, write 5 meta descriptions between 140 and 155 characters. Include a clear next step for the reader. Ban 'rank #1' and fake urgency. Output with character counts." Humans pick one; tools that truncate still need a clean primary line.

H1 vs title pattern: "TASK: Given TITLE_TAG and SERP_NOTE, propose 3 H1 options that match intent but are not identical to the title tag. Cap each H1 at 70 characters. Flag any H1 that drops the primary keyword without a good reason." Use this when CMS title and on-page H1 must differ for clarity.

SERP mimic scrub: "TASK: Audit TITLE_OPTIONS against COMPETITOR_TITLES in SERP_NOTE. Flag options that clone a competitor title too closely. Propose one safer rewrite per flag. Never invent competitor rankings." Keeps your pack distinct without copying the top result word for word.

Paste-ready kits for schema drafts

Article JSON-LD pattern: "TASK: From PAGE_FACTS (headline, author name, datePublished, dateModified, image URL if known, description), draft valid JSON-LD for schema.org Article. Use only fields I supplied. Mark missing fields [NEED FACT] outside the JSON. Do not invent ratings, wordCount, or organization IDs. Output one fenced JSON block." You validate in Google's rich results test or your CMS schema tool before publish.

FAQPage pattern: "TASK: From FAQ_PAIRS I paste (question + answer text already on the page), draft FAQPage JSON-LD. Include only pairs I provided. Ban inventing new questions. Cap answers at the word count already on page. Output one JSON block plus a checklist of pairs included." Pair this with on-page FAQ HTML that matches.

HowTo pattern: "TASK: From HOWTO_STEPS I paste (name + text per step), draft HowTo JSON-LD. Include totalTime only if I supplied an ISO-8601 duration. Ban tools or supplies I did not list. Output JSON plus [NEED FACT] for gaps." Use only when the page is a genuine step guide.

Breadcrumb and Organization pattern: "TASK: From BREADCRUMB_TRAIL (name + URL per crumb) and ORG_FACTS (name, url, logo if known), draft BreadcrumbList and Organization JSON-LD as separate blocks. No inventing parent pages. No inventing social profiles." Engineers paste into the template; you confirm URLs resolve.

Schema audit pattern: "TASK: Audit SCHEMA_DRAFT against PAGE_FACTS and on-page content summary. For each property, reply Keep, Soften, or Remove. Soften means the property overreaches what the page shows. Flag schema types the page does not support. Never add AggregateRating or Review unless I pasted review facts." Run before you ship markup.

Step-by-step SEO prompt workflow

Use one chat thread per artifact and per URL. Dumping a brief and a schema pack into one long thread blurs format rules and risks pasting confidential client metrics into a draft you later publish. The loop below keeps a stable SERP_NOTE sheet while you swap only PAGE_GOAL and TASK. You spend free ChatGPT or PromptMake runs on structure, then human time on SERP checks and a rich-results validation before you publish.

Keep SERP_NOTE and PAGE_GOAL in plain text outside the chat: primary keyword, intent, competitor themes, CTA, banned claims. Mark uncertain items with ?. Update the pack when the SERP or page scope changes. The model should never be your only keyword tool or your CMS.

Step 1: Fill SERP_NOTE and page facts

Write the sheets in plain bullets. Example: "Keyword: chatgpt prompts for seo. Intent: commercial how-to. Competitors: list-style prompt dumps; thin '10 prompts' posts. PAA: best prompts for meta titles; can ChatGPT write schema. Slug: /blog/chatgpt-prompts-for-seo. Word count: 1800-2800. Internal links: SEO content formula guide, PromptMake /text. Ban: rank guarantees, fake volume." Ugly notes beat polished fiction.

Add an explicit fence in every prompt: no rank promises, no invented Search Console numbers, no schema for missing page elements. Tell the model to keep unknowns as [NEED FACT]. Guessing "KD 12" when you mean "I have not checked Labs" creates false confidence in the brief.

Step 2: Draft one artifact, then audit claims

Name the artifact and audience before you paste long notes. Ask for an outline first when the brief is new:

TASK: From ARTIFACT_GOAL and AUDIENCE, propose an outline with headers only. Do not draft body text yet. Flag any header that needs facts missing from SERP_NOTE.

Use that outline to decide what to paste next. Partial SERP notes get [NEED FACT] subsections. Missing PAA questions stay out of the FAQ brief until a human copies them. Then run the full skeleton with OUTPUT_SHAPE matching the outline you approved.

Audit pass: "TASK: Audit DRAFT against SERP_NOTE, PAGE_GOAL, and RULES. For each claim about intent, competitors, titles, schema fields, or outcomes, reply Keep, Soften, or Remove. Soften means the claim overreaches the sources. Propose safer wording. Never add rank promises or new metrics."

Step 3: Lock, validate, and save the wrapper

Optional scaffold: open https://promptmake.net/text, describe "ChatGPT prompts for SEO for content briefs, title tags, and JSON-LD schema with SERP_NOTE and claim fences," generate once, then paste your SERP_NOTE into the returned structure. Guests get about three runs per day; free accounts get about five. Use a run to shape the prompt, then finish in your chat model.

Save the winning prompt next to the artifact name and model label so the next URL reuses the same wrapper. Lock audited text into your brief doc, CMS title fields, or schema template before you start the next artifact so invented metrics cannot leak backward. Publish titles and schema only after you or an SEO lead match every claim to SERP_NOTE and validate JSON-LD in a tester.

Mistakes that wreck SEO ChatGPT drafts

Mistake 1: Asking for "SEO-optimized content" with no intent, SERP themes, or URL plan. The model invents keywords and rank timelines. Lock SERP_NOTE first.

Mistake 2: Using the full article-draft formula when you only need a brief, title pack, or schema kit. Full draft chains solve long-form writing. This library solves planning and markup paste jobs. Keep the libraries separate.

Mistake 3: Letting the model invent keyword volume, difficulty, or Search Console charts. Ban those claims in RULES unless you pasted a sourced figure. Fake metrics create bad prioritization.

Mistake 4: Shipping FAQPage or HowTo schema for content the page does not show. Demand PAGE_FACTS and a schema audit. Mismatched markup wastes engineer time and can trigger rich-result issues.

Mistake 5: Mixing brief hype and schema JSON in one mega-thread. Split artifacts. Reuse SERP_NOTE; change TASK and FORMAT.

Mistake 6: Pasting private Search Console exports, client NDAs, or API keys into a consumer chat without your company AI policy check. Redact. Keep sensitive analytics in approved tools.

Mistake 7: Trusting Instant or Flash as the final judge on schema that must validate. Fast models fit title packs and first brief outlines. Route schema audits to GPT-5.6 Sol, Claude Opus 5, or Gemini 3.1 Pro when JSON goes to production.

Mistake 8: Skipping the human SERP and rich-results check. Fluency on screen is not the same as fidelity to the live SERP or valid JSON-LD. Require Keep / Soften / Remove against SERP_NOTE every time.

Model notes for SEO prompts (mid-2026)

ChatGPT often defaults to GPT-5.5 Instant for fast chat. Instant fits title packs, meta variants, and outline passes when you already locked SERP_NOTE. Keep prompts short: ROLE, TASK, FORMAT, FACT packs, RULES. Skip long chain-of-thought slogans.

GPT-5.6 Sol fits harder edit passes: brief audits against SERP_NOTE, schema drafts that must refuse fake ratings, and title packs that must stay under character caps without clickbait. Give goal, constraints, and format. Drop "think step by step" padding on reasoning-class models.

Claude Sonnet 5 handles long competitor outline pastes you want turned into gap briefs. Claude Opus 5 fits careful schema audits when a wrong AggregateRating would ship to production. Gemini 3.5 Flash fits volume work: many title variants, batch meta lines, and short FAQ question lists from one PAA paste. Gemini 3.1 Pro fits hard reasoning over a thick SERP_NOTE when you need conflict flags against PAGE_GOAL.

Hedge on exact menu names in each vendor UI. They shift. Re-check the model picker when you open a new thread. Prompting split that holds: Instant and Flash get RTF plus a short sample of your house brief format when you need matching headers. GPT-5.6 Sol, Opus 5, and Gemini 3.1 Pro get goal plus constraints plus format, with an explicit refuse-to-invent rule and a [NEED FACT] token. All need your SERP_NOTE in the message. None replace a SERP check, Search Console, or a rich-results validator. None promise rankings.

Build SEO prompts with PromptMake /text

Write the rough ask in plain words: brief vs titles vs schema, and the page type. Open https://promptmake.net/text and generate a structured prompt once. Expect labeled sections you can fill with SERP_NOTE and PAGE_GOAL.

Edit keywords, competitor themes, and schema fields yourself. PromptMake cannot know your SERP or CMS. Paste the filled prompt into Instant or Flash for first drafts, or GPT-5.6 Sol / Opus 5 / Gemini 3.1 Pro for claim and schema audits. Keep free-tier runs for scaffolding, not five synonym retries of the same weak "make it more SEO" ask.

Workflow that sticks: SERP_NOTE, refresh PAGE_GOAL, PromptMake scaffold, fill facts, fast model draft, human SERP check, reasoning model audit, you publish. Store one template per artifact type so you do not rewrite ROLE and RULES from scratch each URL.

FAQ

What are the best ChatGPT prompts for SEO in 2026?

The best ChatGPT prompts for SEO lead with ROLE and claim fences, paste SERP_NOTE and PAGE_GOAL, then demand FORMAT for a content brief, title pack, or JSON-LD draft. Add a Keep / Soften / Remove audit against your SERP notes. Use GPT-5.5 Instant or Gemini 3.5 Flash for first drafts and GPT-5.6 Sol, Claude Opus 5, or Gemini 3.1 Pro for schema audits before you publish.

How do I write an SEO content brief with ChatGPT?

Paste SERP_NOTE with primary keyword, intent, competitor themes, and PAA questions you copied. Ask for Intent statement, Must-answer questions, Outline, Differentiation, Internal links, and [NEED FACT] gaps. Ban rank promises and invented volume. Hand the brief to a writer only after you match every claim to your SERP scan.

How should ChatGPT draft title tags and meta descriptions?

Feed SERP_NOTE and PAGE_GOAL, then ask for 8-10 title options under 60 characters and 5 meta lines between 140 and 155 characters. Require character counts and a ban on clickbait rank claims. Run a SERP mimic scrub against competitor titles you pasted, then pick one title and one meta yourself.

Can ChatGPT write schema markup for my pages?

Yes, as a draft. Paste PAGE_FACTS and only the FAQ or HowTo content already on the page. Ask for Article, FAQPage, HowTo, BreadcrumbList, or Organization JSON-LD that matches those facts. Ban AggregateRating and Review unless you pasted real review data, then validate in a rich-results tester before production.

Do ChatGPT prompts for SEO guarantee higher rankings?

No. These kits speed briefs, titles, and schema drafts. Rankings depend on intent match, content quality, links, technical health, and competition. Put an explicit no-rank-promise rule in ROLE and rewrite any prompt that still invents "rank #1" claims.

Should I use GPT-5.5 Instant or GPT-5.6 Sol for SEO prompts?

Use Instant for title packs, meta variants, and outline passes when SERP_NOTE is already in the message. Use GPT-5.6 Sol for brief audits, schema that must refuse fake ratings, or title packs that must stay inside character caps. Compare both only when you measure quality for a recurring URL workflow.

Can PromptMake help with ChatGPT prompts for SEO for free?

Yes. PromptMake /text turns a rough SEO idea into a labeled prompt you can aim at ChatGPT, Claude, or Gemini. Guests get about three generations per day; registered free users get about five. Fill in your own SERP_NOTE and PAGE_GOAL, then paste into Instant for drafts or GPT-5.6 Sol for audits before you publish.