AI Search and GEO · How to

How to use ChatGPT for SEO: what to hand it and what to keep

Part of the guide Generative engine optimization (GEO): a practical guide for 2026

Use ChatGPT for SEO tasks where it works from text you give it: title and meta variants, outlines, FAQ questions mined from support emails, regex and spreadsheet formulas, and critiques of a draft. Do not use it for search volumes, rankings or facts about other companies unless you verify them. Also use it as a window: run buyers' questions in ChatGPT search, logged out, to see which pages win.

Which SEO tasks to give ChatGPT

Hand it tasks that transform text you supply, and keep tasks that need live data or judgment about facts. Here is a task by task verdict: The strategy these tasks support is in the generative engine optimization guide.

Task Verdict Why
Title and meta description variants Good Pure rewriting within limits you set
Outline from your notes Good Structure is its strength
FAQ questions from support emails Good Pattern-finding in text you paste
Regex, formulas, small scripts Good, then test Fast to write, easy to check
Critique of a draft Good A second reader on demand
First draft of an article Use with care Needs your facts and heavy editing
Keyword search volumes Do not use It has no volume data and may produce numbers anyway
Competitor prices or features Do not trust Often outdated; check their pricing page and date it
Bulk pages at scale Do not do Google's spam policy names it as scaled content abuse

Seven prompts that hold up

Each prompt gives ChatGPT the data and a format, and forbids invention.

1. Title and meta variants

Page topic: [one sentence]. Main query: [query].
Write 8 title tags under 60 characters, the query near the start,
no clickbait, no year unless I give one. Then 5 meta descriptions
of 140 to 160 characters that say what the reader gets. Show character counts.

Check the counts yourself; models often miscount characters.

2. FAQ mining from support emails

Below are 40 customer emails with personal details removed.
List the distinct questions people ask, merged where they mean the same,
with how many emails asked each. Quote the customer's own wording.

Real questions in customers' words make better H2s and FAQs than anything brainstormed.

3. Outline from your notes

Here are my notes, test results and screenshots described in text: [paste].
Build an outline where each H2 is a question a buyer would ask and lists
the fact from my notes that answers it. Mark H2s where my notes have no answer.

4. Draft critique

Read this draft as a skeptical buyer comparing three options.
List: claims without evidence, numbers without dates, places I used
a category instead of a name, and questions you still have at the end.
Do not rewrite anything.

5. GA4 regex

Write a GA4 regex for the Source dimension that matches chatgpt, perplexity,
claude.ai, gemini.google, copilot.microsoft and deepseek. GA4 uses full
match by default, so wrap the pattern so it works as "contains".
Then give 5 sample sources that should match and 3 that should not.

Test the result before trusting it. The setup it plugs into is in AI search traffic in GA4.

6. Internal link suggestions

Here is a list of my URLs with their titles: [paste].
For the draft below, suggest up to 5 internal links: the exact sentence,
the anchor text (descriptive, not "click here"), and the target URL from my list only.

7. Sub-question list for a buyer prompt

A buyer asks: "[realistic prompt]". List the short web searches you would
run to answer it well. Then group them by page type: comparison,
alternative, integration, constraint, feature, problem.

This feeds your fan-out list; verify it by searching the queries yourself.

Use ChatGPT search to study your market

ChatGPT is also one of the places your buyers ask for recommendations, so look at what it answers. Log out or use a temporary chat, ask the question your buyer would ask, and open the sources rather than reading the answer. Three runs per question, because sources change between runs.

What to note: which of your competitors are named, which third-party pages are cited (lists, reviews, forums), and whether any of your pages appear. Those cited pages are where the answer comes from, so they are your outreach and content targets. The full routine, with a template, is in track AI citations for free, and the fixes for missing citations are in how to get cited by ChatGPT.

Make sure ChatGPT can read your site

OpenAI runs three relevant agents, and only one decides whether you can appear in ChatGPT search:

Agent Role (per OpenAI) What to do
OAI-SearchBot Surfaces sites in ChatGPT search results Allow it; blocked sites "will not be shown in ChatGPT search answers"
GPTBot Crawls content that may be used to train models Your choice; independent of search
ChatGPT-User Visits pages when a user's request needs it User-initiated; "robots.txt rules may not apply"

OpenAI says robots.txt changes take about 24 hours to apply.

Mistakes that cost rankings

  • Asking for keyword volumes. ChatGPT has no search volume database. Use Search Console or a keyword tool.
  • Trusting links it gives you. Models can produce URLs that do not exist. Open every link before citing it.
  • Publishing first drafts. Unedited model text is commodity content. Google's AI optimization guide asks for "unique expert or experienced takes that go beyond common knowledge".
  • Publishing at volume. Google's spam policies list "using generative AI tools or other similar tools to generate many pages without adding value for users" as scaled content abuse. On a young domain, a couple of solid pages a week beats dozens of thin ones.
  • Pasting private data. Remove customer names, emails and anything under NDA before pasting.

ChatGPT vs a dedicated SEO tool

ChatGPT is a general assistant; it does not track rankings, crawl your site on a schedule or publish to your CMS. A dedicated tool does one job continuously. If you mostly need help thinking and rewriting, ChatGPT is enough. If you need articles planned, written, published and refreshed every week, use a tool for the routine and ChatGPT for the thinking. A comparison of AI SEO tools covers the options, and how to use Claude for SEO covers longer analysis workflows.

Postbing is one such tool: it plans, writes and publishes articles to WordPress, Shopify, Webflow, Ghost and Notion, and with bring your own key you can run it on your own OpenAI key. Plans start at $29 a month for BYOK Lite as of November 2026.

FAQ

Can ChatGPT do keyword research?

It can suggest topics and group queries, but it has no search volume data and may make numbers up. Use its ideas as a starting list and check volumes and rankings in Search Console or a keyword tool.

Is content written by ChatGPT bad for SEO?

Not by itself. Google's spam policies target pages generated at scale without adding value for users. Edited, fact-checked content with your own information is fine; mass-produced first drafts are the risk.

Allow OAI-SearchBot in robots.txt and your firewall, serve your text in raw HTML, and publish pages that match the searches ChatGPT runs, such as comparisons and alternatives.

Should I block GPTBot?

That is a choice about training, not search. OpenAI says the settings are independent, so you can block GPTBot and still allow OAI-SearchBot to appear in ChatGPT search.

Sources

Last updated: November 16, 2026. Spotted an error? Email support@postbing.com and we will fix it. Editorial policy.

Try Postbing free

Postbing reads your site, plans a month of topics from what your customers search for, writes in your voice and publishes to WordPress, Shopify, Webflow or Ghost. Your first article is free.

Try Postbing free

No card needed. No link exchanges, ever.