How to Use ChatGPT for SEO the Right Way: The 8-Job Playbook

Written by: Arjun Karnik, Growth Marketing Specialist

Key Takeaways

• AI summaries now dominate buyer research, with only 8% of users clicking traditional results when summaries appear, so citation-focused content now drives visibility.
• Generic keyword-stuffed content rarely earns AI citations; a research-first workflow that maps fan-out queries and feeds real Search Console data consistently outperforms it.
• Aligning URLs, titles, and H1s to buyer language, then layering first-hand expertise and FAQPage schema, increases the odds that AI systems can parse and cite your pages.
• Continuous freshness loops and post-publish audits prevent content decay and keep pages citable across ChatGPT, Google AI Overviews, Perplexity, and Gemini.
• Arjun Karnik’s documented test-lab results prove this eight-job system works; see the test-lab results in action to watch how AI Growth Agent runs the loop automatically.

1. Start With Site Goals and Search Console Data

Traditional SEO is losing ground because AI summaries intercept most buyer research before users see ranked results. This playbook solves that problem by focusing your entire strategy on earning citations inside those AI-generated answers. The first job sets context so ChatGPT can diagnose where your site already wins and where it misses buyer questions.

Export your Google Search Console performance report for the last 90 days. Pull queries, impressions, clicks, and average position. Then open ChatGPT and use this prompt template:

I run [describe business in one sentence]. My site currently ranks for the following queries [paste top 20 by impressions]. My click-through rate is declining even as impressions rise. My goal is to earn citations in ChatGPT, Google AI Overviews, Perplexity, and Gemini for buyers researching [describe category]. Identify the three biggest gaps between what I rank for and what a buyer would ask an AI assistant before purchasing in this category.

Before this step, Arjun’s site showed the classic scissors pattern: impressions climbed while clicks fell. The diagnosis was clear: the site ranked for practitioner keywords, while buyers asked different questions in AI assistants. Feeding Search Console data into ChatGPT revealed that gap by highlighting buyer questions with no coverage. Reorienting the content strategy toward those questions rather than keyword clusters meant the GEO subfolder became the only source of new impressions on the entire domain within 60 days, measured in Google Search Console.

Feeding ChatGPT real data such as Search Console exports, competitor URLs, and existing content instead of asking it to guess is one of three core principles for better ChatGPT SEO results, per the Techpresso AI Academy guide. The model’s job in this workflow is analysis and gap identification, not content generation from thin air.

2. Map SERPs and Competitors Across Both Google and AI

The second job maps competitors on both Google and AI answer surfaces. The target is no longer only the traditional SERP. An Ahrefs study of 15,000 prompts found only 12% overlap between URLs cited by ChatGPT, Gemini, and Copilot and Google’s top 10 results for the same queries. Ranking first on Google does not guarantee an AI citation, so your analysis must cover both layers.

Use this prompt template:

For the query “[target query]”, list the types of sources that would be most likely to be cited in a ChatGPT or Google AI Overview answer. Then analyze these competitor URLs [paste 3–5 URLs] and identify: (1) what structured claims they make that are easy to extract, (2) what fan-out questions they answer that I do not, and (3) what first-hand data or expertise they lack that I could supply.

In Arjun’s test lab, this analysis consistently surfaces a structural gap: competitor pages rank on domain authority but lack extractable, one-sentence claims. Kevin Indig’s analysis of 1.2 million ChatGPT responses found that 44.2% of citations come from the first 30% of a page. Competitor pages that bury their core claim in paragraph four are vulnerable to a challenger who leads with it.

3. Turn Competitor Gaps into a Fan-Out Query Map

The structural vulnerability you found in Job 2 becomes your production roadmap in Job 3. Fan-out query extraction is the job that separates a research-first GEO workflow from a volume-based SEO play. A single buyer prompt triggers multiple hidden retrieval queries underneath it. Google AI Mode generates 8–12 sub-queries per prompt (averaging 10.7) while ChatGPT generates 2 to 4, and each one represents a separate retrieval path that content must match to earn citations.

Use this prompt template:

A B2B buyer types this prompt into an AI assistant: “[primary buyer question]”. List every sub-query the AI would need to answer in order to construct a complete response. Group them by intent: informational, comparative, and decision-stage. Then identify which sub-queries have the least published coverage.

In a controlled test on Arjun’s site, pages rewritten to match extracted fan-out queries earned citations while control pages that kept their original keyword-optimized structure did not. A SurferSEO study of 173,902 URLs found that pages ranking for both a main query and its fan-outs are 161% more likely to be cited in Google’s AI Overviews. The fan-out map becomes the production queue for every later job in this workflow.

4. Rewrite URLs, Titles, and H1s in Buyer Language

Buyer-language alignment turns your fan-out map into pages that AI systems can match to real prompts. Jargon creates a retrieval barrier because the machine matches a buyer’s question to the closest available answer. When a page is titled in practitioner vocabulary instead of buyer vocabulary, the match fails even when the content is correct.

Use this prompt template:

Here is a list of my current page titles and slugs [paste list]. Here is the fan-out query map from the previous step [paste map]. Rewrite each title, H1, and slug to match the exact language a buyer would use when asking an AI assistant. Prioritize question-format H1s. Flag any title that uses industry jargon a non-specialist would not search for.

On Arjun’s site, a page titled “What is GEO” was relabelled “How to Get Your Business Recommended by AI Search,” with the slug, title, H1, and H2s all realigned to buyer questions. Citations followed within weeks of that specific change. New articles built on this buyer-language foundation reached thousands of monthly Google impressions within weeks of publication, measured in Google Search Console and contributing to the GEO subfolder growth mentioned earlier.

By 2026, brand visibility in search depends less on page position in ranked results and more on whether a brand is cited within AI-generated responses from systems such as Google AI Overviews and Bing generative search. Buyer-language alignment is the fastest single lever for closing that gap on existing pages.

5. Add First-Hand Expertise and Schema That AI Can Parse

Pages earn citations when they add something new and make that information easy for machines to extract. AI systems cannot cite what they cannot parse, and they rarely cite generic claims that appear everywhere. Two inputs make a page citable: first-hand expertise that introduces specific data, and schema markup that makes the page machine-readable.

Use this prompt template:

Here is a draft section on [topic] [paste draft]. Here is the first-hand data I have from my own work: [paste specific numbers, test results, or observations]. Rewrite the section so that every H2 opens with a direct answer, every claim is supported by a specific number or named example, and the first-hand data appears in the first two sentences of each section. Then generate FAQPage JSON-LD schema for the five most common buyer questions this section addresses.

The Princeton GEO study found up to +115% visibility lift for position-5 pages via attribution, with statistics addition yielding +31–41% and quotation addition +41% on position-adjusted word count. FAQPage schema shows mixed results in isolation: one analysis found 3.2× higher AI Overview citation rates, while a larger 1,885-page study saw minimal impact from schema alone.

The difference lies in how schema is used. In Arjun’s test lab, schema is applied to every published page as a structural requirement, and it always appears alongside first-hand data in the opening sentences. That combination of specific expertise and FAQPage schema produced citations on rewritten pages, while pages with only one of those inputs often remained uncited.

6. Audit Published Pages Before They Enter the Loop

Even perfectly structured drafts can fail if technical or structural issues block AI crawlers. A post-publish audit catches problems that prevent the retrieval layer from reading the page correctly, even when the content itself is strong.

Use this prompt template:

Here is the published HTML of my page [paste HTML or key sections]. Audit it against these criteria: (1) Does the H1 match the primary fan-out query? (2) Does each H2 open with a direct, one-sentence answer? (3) Are all statistics attributed to a named source with a date? (4) Is there FAQPage schema present? (5) Are AI crawlers permitted in robots.txt? Flag every failure and provide a corrected version of each failing element.

Pages with definitive headings, concise explanations, comparison sections, schema support, and evidence-backed statements are dramatically easier for AI systems to parse, summarize, and cite than dense, unfocused content. The post-publish audit enforces those standards before a page enters the freshness loop.

On Arjun’s site, this step caught blocked AI crawlers on several pages that had been live for months. Unblocking those crawlers was a prerequisite for any citation work downstream. Technical plumbing forms the foundation that every other job relies on.

7. Build Internal Links and a Self-Healing Freshness Loop

Content freshness keeps citations from decaying after you earn them. Content decay is the silent killer of citation performance. In Arjun’s tests, pages dropped 78% to 99% in two months without updates, and Seer Interactive’s analysis of 47,097 AI citations across 7,683 pages confirmed the pattern at scale.

Their study found that 75% of cited pages had been updated within the last year, with pages cited consistently across all four months averaging under six months since their last update. The implication is clear: freshness is a structural requirement for staying citable, not a nice-to-have maintenance task.

Use this prompt template:

Here is my current content library with publication dates and last-updated dates [paste list]. Here are my Search Console impression trends for each URL over the last 60 days [paste data]. Identify which pages show impression decay of more than 20% month-over-month. For each decaying page, suggest three internal linking opportunities from higher-performing pages, and list the five specific updates—new statistics, added FAQ items, refreshed examples—that would most improve citation probability.

Via AI Growth Agent, Arjun’s system runs impression-decay tripwires that automatically queue updates when a page’s performance drops below threshold. This produces self-healing content that repairs itself on a loop instead of waiting for a quarterly audit. The system runs at 5 to 8 autonomous actions per day, combining new articles with updates to existing ones, via AI Growth Agent. Approximately 50% of sources cited for a given prompt will change within 13 weeks, which makes continuous refresh a structural requirement rather than a best practice.

See the self-healing content system in action to watch the freshness loop run automatically without manual audits.

8. Run Defensive GEO Audits in Parallel With Growth

Defensive GEO protects your brand narrative while the growth loop runs. AI systems already have answers about most businesses, and some of those answers are wrong. A wrong AI answer causes more damage than no answer because it reaches buyers at the exact moment they form vendor preferences.

Use this prompt template:

Ask ChatGPT, Gemini, and Perplexity each of the following questions about my business [list 5–10 buyer questions]. Record the exact answers. Then compare those answers against these verified facts about my business [paste accurate information]. Identify every factual error, missing capability, or competitor misattribution. For each error, draft a corrective content block—a self-contained paragraph with a direct claim, a supporting statistic, and a named source—that can be published on my site to create a machine-readable correction.

In Arjun’s test lab, the defensive GEO audit runs in parallel with growth work, not after it. The visibility audit across ChatGPT, Google AI Overviews, Perplexity, and Gemini is the first step before any new content goes live, because correcting the existing record creates a clean baseline. AI Growth Agent clients average more than 12,000 additional AI citations and mentions, over 100,000 additional bot visits, and a 20% or greater lift in impressions across the first twelve weeks. That outcome depends on accurate machine-readable facts, which the defensive audit establishes.

Frequently Asked Questions

Can ChatGPT do SEO?

ChatGPT can perform specific, well-defined SEO jobs when given real data to work with. It can extract fan-out queries from a buyer prompt, rewrite titles and H1s to match buyer language, generate FAQPage schema, audit published HTML for structural problems, identify internal linking opportunities, and flag content that is decaying based on Search Console data. It cannot crawl the web in real time, access live search volume data, or monitor rankings autonomously. The correct use of ChatGPT in an SEO workflow is as an analysis and structuring tool fed with real site data, not as a content generator operating from a blank prompt.

Is ChatGPT good for SEO?

ChatGPT is effective for SEO when you place it inside a research-first workflow that feeds it real inputs. The documented outcomes from Arjun Karnik’s public test lab show that pages rewritten using fan-out queries extracted from ChatGPT earned citations while control pages did not, and that new articles built on buyer-language alignment reached thousands of monthly Google impressions within weeks. ChatGPT is not effective when used to generate generic content at volume without human expertise, first-hand data, or schema markup layered in. Google does not penalize AI-assisted content as a category; it penalizes content that lacks originality, expertise, and helpfulness regardless of how it was produced.

Is SEO dead with AI?

SEO is not dead, but the target has shifted. Traditional SEO focused on rankings on a human-readable list of ten blue links. The same technical foundations—structured content, internal linking, crawlability, and topical authority—now serve a second surface: AI-generated answers. On Arjun Karnik’s site, articles built for citation in AI answers also reached thousands of monthly Google impressions within weeks, and the GEO subfolder became the only source of new impressions on the domain within 60 days. Businesses that lose treat AI search as a separate channel requiring a separate strategy. Businesses that win apply a unified research-first workflow that earns both Google impressions and AI citations from the same content.

What is the difference between SEO and GEO?

SEO optimizes for ranking in search engine results pages to earn clicks. GEO—generative engine optimization—optimizes for being cited inside AI-generated answers from ChatGPT, Google AI Overviews, Perplexity, and Gemini. The authority model differs: SEO earns authority through backlinks and domain authority, while GEO earns it through topical coverage and structured, extractable content. The query model differs: SEO targets the keyword the buyer typed, while GEO targets the dozens of fan-out queries an AI assistant generates underneath a single buyer prompt. The success metric differs: SEO tracks rankings and clicks, while GEO tracks citations, mentions, and share of answer. The freshness requirement differs: SEO content can hold a position for months without updates, while in Arjun Karnik’s tests, pages dropped 78% to 99% in citation performance within two months without maintenance.

How long does it take to earn AI citations?

Timelines depend on your starting point and how tightly you follow the workflow. On Arjun Karnik’s site, buyer-language relabelling of a jargon page produced citations within weeks of the change. New articles built on fan-out query mapping reached thousands of monthly Google impressions within weeks of publication. The general pattern observed across the test lab is coverage and impressions in weeks, citations in one to three months, and compounding topical authority after month three. Perplexity, which uses real-time retrieval, reflects content changes within hours or days. ChatGPT, which relies primarily on training data updated periodically, takes longer to reflect new content. Running the full eight-job workflow via AI Growth Agent at 5 to 8 autonomous actions per day accelerates the timeline by maintaining both freshness and structural quality at the same time.

Conclusion: One Loop for Both Google Impressions and AI Citations

The eight jobs in this playbook operate as one loop that turns raw data into citable, visible content. You feed real data in, extract fan-out queries, align language to buyers, embed expertise and schema, audit after publishing, trigger the freshness cycle, and correct what AI already says. Each job hands structured inputs to the next, so the system compounds instead of resetting.

The documented outcome from Arjun Karnik’s public test lab is that this loop works on both surfaces simultaneously. The GEO subfolder became the only source of new impressions on the entire domain in 60 days. Pages rewritten to match fan-out queries earned citations while controls did not. Buyer-language relabelling produced citations within weeks. The content that earns AI citations is the same content that earns Google impressions, because both surfaces reward structured, fresh, specific, expert content aligned to buyer questions.

Cadence, not knowledge, is the main constraint. One person cannot publish and refresh at the volume and frequency this channel now requires. The client outcomes described earlier—12,000+ citations, 20%+ impression lifts—reflect the advantage of running this loop at machine cadence. Arjun’s system runs at 5 to 8 autonomous actions per day via AI Growth Agent, combining new articles with updates on autopilot so founder time no longer blocks growth.

The window for outsized gains is open now because answers gain incumbency. Businesses that decode the new answer layer first own the category narrative before competitors adapt. See how AI Growth Agent runs the full eight-job loop for your site and turns it into measurable citation and impression wins for your business.