Generative AI Search Engine Optimization Agency Work: Which AI Models Draft It Best, With Live GoldieBench Scores

A generative AI search engine optimization agency still does the strategy, but AI models now draft much of the page, brief and markup work underneath it, and some models are measurably better at that than others.

This page matches those SEO tasks to models using only scores read on live GoldieBench pages on 11 November 2026, each quoted with the page it came from.

For the parts that need people, book a free SEO strategy session with my team.

Why model choice matters less than eligibility

Google's guide to generative AI features says AI Overviews and AI Mode are rooted in its core ranking and quality systems, retrieving pages from the Search index before generating an answer.

OpenAI's bot documentation says ChatGPT search only shows sites that allow OAI-SearchBot.

No model fixes a blocked crawler or a thin page. A good model only speeds up the drafting once the basics are right.

Keep that order in mind before reading the scores.

How GoldieBench scores models

The GoldieBench models page says each model gets the same fixed prompts, one shot each, scored 0–10 by Julian.

The leaderboard lists 50 tasks, 33 models and 1310 live demos, split into Games, Visuals, Sims and Pages.

Pages is the relevant category for SEO work, and the task index lists three Pages tasks: Landing, AIPB Promo and Web-OS Desktop.

There is no GoldieBench task for briefs, schema or audits, so those picks lean on overall strength.

Landing task: the nearest test for service pages

The Landing task page asks for a modern marketing landing page in one HTML file, with no iteration.

It reports 26 models tested, 21 of 26 scored, an average of 8.11/10 and Fugu Ultra as the winner.

Four models tie at 9.0, so a single task can't separate the leaders on its own.

Checking consistency with a second Pages task

The AIPB Promo task page scores an animated promo build: 26 models tested, 20 of 26 scored, an average of 7.48/10, with Qwen 3.8 as the winner.

Comparing both tasks shows which models hold their level across different page briefs.

GPT-5.6 Sol and Claude Fable 5 have the smallest gaps, while Grok drops the most between the two.

For agency work, where every client brief differs, a small gap is worth more than one top score.

Overall leaderboard positions for the shortlist

On the GoldieBench leaderboard, ranked by average score across all tasks:

• Fusion is #1 at 8.59, best at Games.
• Claude Opus 5 is #2 at 8.27, best at Sims.
• GPT-5.6 Sol is #4 at 8.16, best at Pages.
• GLM-5.2 is #11 at 7.77, best at Pages.
• Gemini 3.6 Flash is #18 at 7.08, best at Pages.
• Claude Sonnet 5 is #19 at 7.01, best at Pages.

GPT-5.6 Sol is the highest-ranked model whose strongest category is Pages.

Which model for which agency task

Treat the table as a starting point and swap models if your own drafts disagree.

A drafting workflow that keeps the scores fair

A benchmark score tells you how a model performs on a fixed prompt, not on your client's brief.

Run a small test of your own before you commit a model to agency work.

1. Pick one real service page that needs rewriting.
2. Give the same brief to two shortlisted models, for example GPT-5.6 Sol and Gemini 3.6 Flash.
3. Mark every factual claim in both drafts and check each against the business owner's notes.
4. Count how many claims you had to delete or correct.
5. Keep the model that needed fewer corrections, even if the other draft looked better.

Repeat the test every few months, because the leaderboard moves as new models arrive.

Log the results in a shared sheet so the whole team picks models from evidence rather than habit.

What the scores can't tell you

• Whether a page is accurate. GoldieBench scores builds, not facts.
• Whether a page is non-commodity. Google's guide says first-hand experience matters most, and no model has yours.
• Whether you'll be cited. Bing's Webmaster Guidelines say GEO does not guarantee grounding or citations.
• Whether a model's schema is valid for your business. Google says structured data should match the visible page.

Google's spam policies also warn that mass-produced pages made mainly to manipulate rankings count as scaled content abuse, however they are created.

Prompt hygiene that lifts any model

The same model writes very different pages depending on the brief it gets.

• Paste in the owner's own notes, so the draft starts from first-hand facts.
• List the buyer questions the page must answer, in the order a buyer asks them.
• Ban invented numbers, reviews and client names in the instructions.
• Ask for plain HTML headings and short paragraphs, so a person can edit quickly.

Good inputs matter whichever model you pick, because no score measures your brief.

The human and agency part

Models draft. People decide which pages exist, what they claim and who links to them, so Book a free SEO strategy session if you want that part handled.

Google's hiring guidance says no one can guarantee a #1 ranking, and OpenAI says ChatGPT search placement is not guaranteed.

Julian Goldie's agency, Goldie Agency, handles SEO, GEO, AEO and links, and calls its outreach "manual outreach, no PBNs".

Julian's self-reported figures list 1,000+ campaigns and 240+ clients, and his YouTube channel has 425K+ subscribers.

Want to run these models yourself? The AI Profit Boardroom teaches AI SEO step by step, and for the ChatGPT side read our page on how to rank in ChatGPT.

Scores tell you which model drafts best, not what to publish. Book a free SEO strategy session for that part.

Also On Our Network

• Agent Operatorsthe seven-step operator process, with robots.txt lines and snippet-tag checks
• AI Income Deskwhether a small business should hire one, with five business-type playbooks
• AI Tool Verdictfive ways to get the work done, each scored out of 10
• agentos.guidefive draft-only Agent OS prompts that prepare the audit before you hire

Pick a consistent Pages scorer for drafts, keep people on facts and links, and book a free strategy session when you want a generative AI search engine optimization agency to handle the rest.