AI SEO Agency London Tasks And The AI Models To Use For Them, With Real GoldieBench Landing And Leaderboard Scores

Any AI SEO agency London clients hire now drafts briefs, schema and landing pages with AI models, and the model chosen changes how much human editing each page needs.

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

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

What the bench can and cannot tell an SEO team

The GoldieBench models page sets out the method: the same fixed prompts, one shot each, scored 0–10 by Julian.

The leaderboard covers 50 tasks and 33 models, with 1,310 live demos.

The tasks page groups them into games, sims, visuals, pages and a few other tests, and each one asks for a single HTML file.

Only the page tasks look like marketing work, so they are the closest signal for local landing pages.

No GoldieBench task scores schema markup, keyword briefs, Business Profile checks or outreach research.

The London SEO tasks a model touches

Landing page scores: the nearest match for borough pages

The Landing task page asks each 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 as the winner.

A polished layout is only the shell of a borough page.

The page still needs your real jobs, photos and answers, which no benchmark measures.

Overall averages: who holds up across every task

The GoldieBench leaderboard ranks models by their average score across all tasks.

GPT-5.6 Sol is the highest-ranked model whose best category is Pages, which makes it a sensible first pick for page drafts.

GLM-5.2 and Gemini 3.6 Flash are also listed as best at Pages, with lower overall averages.

Pick one model for page drafts and stick with it for a month, so your editors learn its habits.

Switching models every week makes it harder to tell whether a weak page came from the model or from a thin brief.

Score your own drafts the same way the bench does: same brief, one attempt, a 0–10 mark from the person who edits it.

Matching models to each London job

• Borough and service pages: start with a Pages-strong model such as GPT-5.6 Sol (8.7/10 on Landing), or a 9.0 scorer such as GLM-5.2 or Fugu Ultra, then edit every claim.
• Schema drafts: no GoldieBench score covers this, so any capable model will do, as long as each value matches the visible page, as Google's LocalBusiness guide and AI features page expect.
• Briefs from Search Console: no score covers this either; use the model your team trusts and check it invents nothing.
• Crawler audits: the agent's ability to fetch robots.txt and headers reliably matters more than the model.

What no model fixes for a London business

A model cannot open your server to the right bots if the firewall still blocks them.

OpenAI's crawler overview says sites opted out of OAI-SearchBot will not appear in ChatGPT search answers, and Perplexity's docs say the same for PerplexityBot.

A model also cannot make a co-working desk eligible for a Business Profile; Google's guidelines require signage, staff and customer hours.

And it cannot earn genuine reviews; Google's Maps contribution policy bans incentives and review gating.

How to read the gaps between scores

On the Landing task, four models share 9.0/10, and the next two sit at 8.7 and 8.6.

For a borough page, a gap that small matters less than the brief you give the model.

The bigger gaps, such as 9.0 against 6.5, are where a model choice really shows up in editing time.

Remember each score is a single one-shot attempt, so treat a tie as a tie.

A prompt pattern for London page drafts

Whichever model you pick, give it the facts rather than asking it to find them.

• The service, the borough and the top search query for that pair.
• Three real recent jobs in that area, with permission to mention them.
• Your opening hours, phone number and service area, exactly as on your Business Profile.
• An instruction to open with a direct answer and to invent nothing.

That pattern keeps the draft consistent with your profile, which matters because Google's Business Profile guidelines want your real-world name and details used consistently.

The UK caveat on ChatGPT Business Hub

OpenAI's help centre says (Nov 2026) that Business Hub lets businesses claim a ChatGPT listing, but for now only existing U.S. listings are open to local shops and restaurants.

So for a London café or shop, the website and Google profile remain the main levers, whatever model drafts them.

The human and agency half of the job

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

So the work that decides results is still human: picking the pages, checking facts, outreach and reading the reports.

If you want that half handled, book a free SEO strategy session and bring your service list and the boroughs you cover.

Goldie Agency works with London businesses remotely, and by Julian Goldie's own published numbers it has run 1,000+ SEO and link building campaigns for 240+ clients.

Its links come from manual outreach, no PBNs, with paid placements marked rel="sponsored" under Google's link spam policy.

If you would rather do the human half yourself, learn AI SEO yourself inside the AI Profit Boardroom.

A model-assisted example: a Peckham bakery

1. GPT-5.6 Sol drafts a page for "sourdough bakery Peckham", and the owner adds the real bakes and opening times.
2. Any capable model drafts Bakery schema from the contact page, and the owner checks each value.
3. An agent confirms OAI-SearchBot and PerplexityBot are allowed.
4. The owner asks every customer for a review with the same link.
5. Each month, the owner checks Search Console and Bing's AI Performance report.

Also On Our Network

• Agent Operatorsthe eight-step process I would run on a London business site
• AI Income Deskwhether hiring one is worth it for a London small business
• AI Tool Verdictfive routes to London AI search visibility, rated out of 10
• agentos.guidefive Agent OS prompts that do the London groundwork before you hire

Use a Pages-strong model for drafts, keep people on facts, profiles and outreach, and book a free strategy session with the AI SEO agency London businesses work with remotely.