Outsource SEO Services Or Draft In-House? Live GoldieBench Scores For The Models Behind Briefs, Schema And Landing Pages

When you outsource SEO services, part of the job is drafting: content briefs, schema markup and landing pages, and AI models now do a lot of that first draft.

This page shows which models score well on the closest GoldieBench tasks, using only scores read on live GoldieBench pages on 11 Nov 2026.

For the part no model can do, book a free SEO strategy session with my team.

What GoldieBench tests, and where SEO fits

The GoldieBench leaderboard lists 50 tasks and 33 models across four categories: Games, Visuals, Sims and Pages.

Its rankings use Julian's 0–10 scores averaged across every task a model ran.

Pages is the only category close to SEO work, and the leaderboard shows 3 tasks in it.

There is no GoldieBench task for content briefs, schema or outreach research, so for those jobs the scores are a loose guide, not a verdict.

The SEO tasks behind outsourcing, and the benchmark signal for each

Landing task scores

The Landing task page asks every model for a modern marketing landing page in one shot.

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

For service pages, those top four are a strong place to start a first draft.

A great-looking draft still needs your real services, locations and proof before it goes live.

AIPB Promo task scores

The AIPB Promo task page sits in the same Pages category and asks for a cinematic promo for a real offer.

It reports 26 models tested, 20 of 26 scored, an average of 7.48/10, and Qwen 3.8 as winner.

GPT-5.6 Sol and Gemini 3.6 Flash rank highly on both Pages tasks, which makes them steady picks for page drafts.

Overall leaderboard: who holds up across every task

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

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

For mixed SEO work, a high overall average suggests a model that copes well with varied prompts, which helps when one agent drafts pages, schema and briefs in a row.

Why one-shot scores matter for SEO drafting

The Landing page says every model got the exact same prompt inside the Agent Operating System, produced a single HTML file, and had no iteration and no examples.

That is close to how a busy owner or agency uses a model: one clear instruction, then a draft to review.

A model that scores well in one shot usually needs fewer rounds of fixes before a page is ready for a human edit.

It does not mean the facts are right, and none of these tasks checks claims, local details or compliance.

A drafting workflow to pair with an outsourced team

1. Your provider writes the brief: target query, search intent, the questions buyers ask and the proof you can show.
2. A Pages-strong model turns the brief into a first draft and a matching schema draft.
3. Someone who knows the business replaces every generic line with real detail, photos and examples.
4. The provider checks headings, internal links, schema and crawl access, then logs the change.
5. You approve the page before it goes live, and the provider reports on it against your baseline.

Want to run steps 2 and 3 yourself? The AI Profit Boardroom is where you learn AI SEO yourself with the same models.

Matching models to the drafting jobs

• Service and landing pages: start with a Landing 9.0 scorer such as Fusion, GLM-5.2 or Grok, or with GPT-5.6 Sol, then edit for accuracy.
• Promo-style pages for offers: GPT-5.6 Sol posted the strongest listed AIPB Promo score at 8.3/10.
• Schema drafts: no benchmark exists, so use any capable model and check every value against Google's Organization structured data guide.
• Content briefs: no benchmark exists either, so use a model you trust and ban invented facts in the prompt.

What no model handles for you

A model can't vet a provider with Google's hiring questions, including the 2026 addition asking whether AEO or GEO advice matches Google's generative AI guidance.

It can't own your Search Console, Business Profile or domain, and it shouldn't.

Keep those in your company's name and give providers named, limited access, starting with read access during an audit, as Google suggests.

It can't earn editorial links either. Under Google's link spam policy, paying for a link that passes ranking credit is a violation; a paid spot needs a sponsored or nofollow attribute.

And it can't open your site to AI search: OpenAI's crawler overview says blocking OAI-SearchBot keeps you out of ChatGPT search answers, and its help article says placement is not guaranteed.

The human and agency part

Outsourcing works best when models do first drafts and people handle strategy, facts, approvals and outreach.

At Goldie Agency the people side spans SEO, GEO and AEO, plus link building pitched as manual outreach with no PBNs.

The agency homepage claims "10+ years of results in tough niches", and Julian's about page lists 425K+ YouTube subscribers.

Like every provider, it can't guarantee a #1 ranking; Google says nobody can.

Small-business example: a physio clinic splitting the work

1. Draft treatment pages with a Landing 9.0 model, then a physio checks every clinical claim.
2. Draft LocalBusiness schema and confirm the address and hours against the site.
3. Keep Business Profile ownership and add the agency as a manager.
4. Hand outreach and technical fixes to the agency, with a weekly change log.

Free extras: the 200+ AI SEO Prompts for briefs and audits, and the free link building book. Or book a free SEO strategy session.

Also On Our Network

• Agent Operatorsthe nine operator steps for running an outsourced SEO engagement safely
• AI Income Deskwhether outsourcing SEO is worth it for your type of small business
• AI Tool Verdictfive routes to getting SEO done, each scored out of ten
• agentos.guidefive Agent OS prompts to run before you hire anyone for SEO

Let the top models draft, keep people on facts and links, and book a free strategy session when you decide to outsource SEO services.