ChatGPT SEO Agency Work + Which AI Models To Use, With Real GoldieBench Page Scores

A ChatGPT SEO agency now leans on AI models for briefs, schema drafts and landing pages, so the model you pick shapes the work before a human ever edits it.

GoldieBench does not score SEO directly, but its Pages tasks come closest to the landing-page work this job needs.

Every number below was read on live goldiebench.com pages on 11 November 2026.

If you want people, not models, on your site, book a free strategy session with Julian's team.

What GoldieBench Measures, And What It Leaves Out

The GoldieBench methodology says each task is one fixed prompt, run once, and the output must be a single self-contained HTML file.

Scores are Julian's own 0 to 10 numbers, and the ranking is the average across every task a model was scored on.

The categories are Games, Visuals, Sims and Pages.

Nothing on the bench tests keyword research, citation accuracy or how ChatGPT picks sources.

Treat it as a signal of how well a model ships a working page in one go.

The leaderboard also notes that models without scored verdicts stay unranked, so a missing model is not a low score.

The Models Tagged For Pages

The leaderboard tags each model with the category it does best in.

These are the overall averages it showed, with matching figures on the models page.

Claude Opus 5 is tagged for Sims, but it sits second on the whole board, so it stays in the comparison.

GLM-5.2 and Claude Sonnet 5 also carry the Pages tag, though we only read per-task scores for the three models below.

Fusion tops the overall board at 8.59, but it is a multi-model panel rather than a single model, so we left it out of the page comparison.

Scores On The Three Page Tasks

Each model page lists its per-task scores.

These are the three Page tasks, as shown on the GPT-5.6 Sol, Claude Opus 5 and Gemini 3.6 Flash pages.

GPT-5.6 Sol leads the Landing Page task at 8.7.

Claude Opus 5 leads Webos Page at 8.7.

Gemini 3.6 Flash lands 8.6 on Landing Page, close behind Sol.

Reading The Gaps Between Scores

The spread on the Page tasks is narrow.

Sol, Opus 5 and Gemini 3.6 Flash all land between 7.8 and 8.7 across the three tasks.

That tells you any of them can produce a usable first draft of a web page in one shot.

The bigger differences show up in the overall averages, where Opus 5 at 8.27 sits well ahead of Gemini 3.6 Flash at 7.08.

For long, multi-step SEO work such as a site-wide brief, that wider strength matters more than a single page score.

For quick, repeatable drafts, the Flash model's near-top Landing Page result is the more useful number.

Matching Models To ChatGPT SEO Jobs

Here is how we would split the work, based only on those scores.

Route the heavy thinking to one model and the bulk drafting to another, then put a person on every final check.

Keep a single facts file that every model reads, so the business name, services and service area never drift between pages.

If you want to set up that routing yourself, learn AI SEO yourself with the Boardroom's Agent OS trainings.

Why The Model Matters Less Than Access

No model fixes a site ChatGPT cannot reach.

OpenAI's web search help page says eligibility needs OAI-SearchBot crawl access and a host or CDN that allows OpenAI's published searchbot IPs.

The OpenAI crawler overview keeps that bot separate from GPTBot, the training crawler, so you can allow one and block the other.

The same help page says placement is not guaranteed, whichever model wrote the page.

Prompt Pattern For A Landing Page Draft

Whichever model you pick, give it a tight brief.

Write an answer-first service page for: QUESTION.
Open with a two-sentence direct answer.
Use only facts from facts.md; mark gaps as [NEEDS SOURCE].
Add one comparison table and three FAQs.
Output clean HTML with plain H2s. Do not invent reviews or numbers.

Google's AI optimization guide asks for non-commodity content, so feed the model your own notes, photos and case details.

Its spam policies also class mass-produced pages made to manipulate rankings as scaled content abuse, however they are made.

A Small-Business Example

A three-chair hair salon wants to show up when people ask ChatGPT for a colorist nearby.

1. The owner confirms robots.txt allows OAI-SearchBot.
2. Claude Opus 5 drafts a brief from the stylists' notes on balayage and color correction.
3. GPT-5.6 Sol turns the brief into a landing page with a short answer up top.
4. Gemini 3.6 Flash drafts LocalBusiness schema and three FAQs from the same facts.
5. The owner checks every claim, adds real photos and publishes.

Google's LocalBusiness markup docs cover the schema fields, and the owner keeps them matched to the visible page.

Where The Human Or Agency Part Starts

Models draft fast, but earning mentions on sites ChatGPT cites is relationship work.

Editors respond to people with a useful idea, not to a template, and that part does not show up on any benchmark.

Someone also has to decide which questions matter to the business and which pages are worth building at all.

Google's hiring advice warns that a top ranking can never be promised, so hold any ChatGPT pitch to that same bar.

Goldie Agency, Julian Goldie's SEO, GEO, AEO and link building agency, describes its link building as manual outreach, no PBNs.

Julian's own published numbers include 1,000+ SEO and link building campaigns and 425K+ YouTube subscribers.

Book a free SEO strategy session to see how the team would pair models with outreach for your site.

How To Check The Results

From what we see, ChatGPT referrals usually land in analytics with chatgpt.com as the source, so segment that traffic.

Bing's AI Performance report shows how often your pages are cited in Copilot answers.

For more on the content side, see our post on how to rank in ChatGPT.

To go deeper, learn AI SEO yourself inside the Boardroom, or Book a free SEO strategy session for a team plan.

Free extras: the 200+ AI SEO Prompts pack and the free link building book, Link Building Mastery.

Models can draft the work, but a strategy still needs a human call. Book a free SEO strategy session to talk it through.

Also On Our Network

• Agent Operatorsthe eight steps an operator runs to get a site cited in ChatGPT search
• AI Income Deskwhether a small business should hire for ChatGPT search, and what to ask
• AI Tool Verdictfive ways to get into ChatGPT search, scored out of ten
• agentos.guidefive draft-only Agent OS prompts that do the ChatGPT SEO groundwork

Pick GPT-5.6 Sol for pages and Claude Opus 5 for briefs, keep a human on facts, and book a free strategy session when you want a ChatGPT SEO agency to handle the rest.