JoinLaunched today · 24 July 2026 · Claude Opus 5 × Agent OS SEO
The 24-Hour Citation Machine.
Claude Opus 5 dropped this morning — near-Fable intelligence at half the price. This guide shows the Agent OS workflow it now powers: an AI-written GEO page that Google's AI Overview picked up and recommended as the #1 agency in under 24 hours. The real search, the real page, and a fresh Opus 5 build are all embedded below.
the whole machine in one line — the proof for every stage is embedded in this guide
<24h
from publish to #1 AI Overview citation
107.9s
for Opus 5 to write a full GEO page
$5/$25
Opus 5 per million tokens — half Fable 5
Day 1
already on your Claude subscription
✦
I · the problem
The Invisible Business Problem.
Search changed while everyone was watching their rankings.
Ask Google for a service now and an AI Overview answers first.
It names ONE recommended provider.
Maybe two.
Everyone else on page one is under it — and a growing share of searchers never scroll past the answer.
So you can rank #4 on Google and still not exist.
The AI said someone else's name.
That's the game now: not ten blue links — one sentence, one winner.
The 24-Hour Citation Machine breaks that cycle for good.
Thinking it?"Isn't GEO just SEO with a new name?"
They overlap, but the target is different. SEO competes for a position on a results page. GEO competes for the sentence inside the AI's answer — the entity it names, the source it cites.
The proof below is a page built for that sentence — and it got it in under a day.
✦
II · how it works, in simple words
Write the page the AI wants to quote.
AI Overviews don't invent answers. They lift them.
When someone asks "best GEO agency in Collin County", Google's AI scans for a page that already answers exactly that — a direct answer at the top, clear definitions, an FAQ, local proof.
If your page is the cleanest answer available, the AI quotes YOU.
So the workflow is simple: find questions the AI already answers, then give it a better page to lift. That's it.
And as of this morning, the writing engine got a serious upgrade.
Claude Opus 5 launched today — Anthropic's fourth Claude 5 model. Near-Fable 5 intelligence at half the price ($5 in / $25 out per million), an effort toggle so you choose speed vs depth per task, and — Anthropic's own words — "much stronger at verifying its work and iterating carefully until it succeeds." That last line is exactly what SEO content needs.
the shift in one picture — GEO aims at the sentence, not the list
Thinking it?"Do people actually skip the links now?"
Watch anyone under 30 search. The AI Overview answers, they read it, done — the links are a fallback, not the destination.
That's why the screenshot below matters more than any ranking report: the AI names ONE agency by name.
✦
III · exactly how it works — the run that proved it
Published July 15. Cited #1 the next day.
Here's the exact run, step by step, with the receipts.
1
Pick a question the AI already answers. "GEO Agency in Collin" — a local, commercial query where Google shows an AI Overview and no incumbent owns the answer. That's the whole targeting secret: go where the AI is already talking but nobody's given it a great source.
2
Write the citation page with Claude inside the Agent OS. One prompt produces the full page: a direct-answer opening the AI can lift word-for-word, a clean GEO definition, bulleted services, authority proof, a GEO-vs-SEO table, honest pricing, and a 5-question FAQ with FAQPage schema. Below, you'll watch Opus 5 do this job in 107.9 seconds.
3
Publish it on the real site. juliangoldie.com/best-geo-agency-collin-tx went live July 15 — screenshot below, click through and read it.
4
Push it to the index — don't wait to be found. Submit the URL for indexing the moment it's live. Waiting for a natural crawl is where "same day" becomes "next month".
5
The AI Overview picks it up. Less than 24 hours later, searching "GEO Agency in Collin" returns Google's AI Overview recommending Goldie Agency by name — citing the page. It also ranks organically right under the answer.
6
Multiply. The page is a template. Next county, next city, next service — same machine, another citation. Opus 5 makes each copy in ~108 seconds.
the timeline that used to take a quarter — compressed into a day
IV · the sources — click everything
The receipts, live and official.
"A thoughtful and proactive model that comes close to the frontier intelligence of Claude Fable 5 at half the price… much stronger at verifying its work and iterating carefully until it succeeds."
— Anthropic, the Claude Opus 5 announcement, 24 July 2026
What you're looking at: the live Google result. The AI Overview names Goldie Agency as "the top Generative Engine Optimization agency recommended for Collin County" — and the citation chip is the page from this workflow, published the day before. That's the whole game in one screenshot.
What you're looking at: the page the AI lifted — live on juliangoldie.com. Notice the shape: a direct answer in the first sentence, a plain GEO definition right after, a key-takeaways box. Built to be quoted, not just ranked.
"Rank #1" used to mean a blue link. Now it means the AI says your name.
V · my launch-day test — opus 5 runs the machine
I gave Opus 5 the same job this afternoon.
Opus 5 shipped this morning. By this afternoon it was running the citation workflow inside the Agent OS — on your existing Claude subscription, no new key.
The job: build the next county's citation page — "Best GEO Agency in Denton County, TX" — same structure that won Collin.
High effort: 107.9 seconds. One prompt, one complete 21,537-character page: direct-answer box, entity definitions, GEO-vs-SEO table, honest pricing, five FAQs with FAQPage schema, 34 natural local mentions across Denton, Lewisville and Flower Mound. Zero console errors. It's embedded live below — scroll it.
Opus 5's build · one prompt · 107.9s · scroll inside
What you're looking at: the actual HTML Opus 5 returned, embedded unedited. The blue direct-answer box at the top is the sentence built for the AI to lift — the same pattern the Collin page used to win its citation.
Then I tested the headline feature — the effort toggle. Same prompt, effort set to low: 111.5 seconds, 22,288 characters. For straight content generation, low effort matched high almost exactly.
That's the honest finding, and it's good news for your token bill: run content generation on LOW effort (same page, cheaper), and save HIGH effort for the jobs that need real reasoning — site audits, cannibalization analysis, competitive research.
my launch-day measurement — bars show total output, both runs produced a complete page
Thinking it?"Won't AI-written pages get penalised?"
Google doesn't penalise AI writing — it penalises USELESS writing. The Collin page answers the exact question, defines its terms, shows real proof and prices honestly. That's why the AI lifted it.
The machine writes the draft in two minutes; the honesty in it is the ranking factor.
✦
VI · the framework
The five gears of the 24-Hour Citation Machine.
i.
The Answer Hunt
Find queries where an AI Overview already answers but no strong source owns it. You're not creating demand — you're volunteering to be the citation.
ii.
The Citation Page
Opus 5 writes the page AI wants to lift: direct answer first, clean definitions, FAQ schema, local signals, honest proof. ~108 seconds per page.
iii.
The Fast Lane
Publish and push to the index the same hour. The clock starts at submission, not at "whenever the crawler wanders by".
iv.
The Lift
The AI Overview quotes your page — by name, with your citation chip. One sentence that outranks every blue link under it.
v.
The Multiplier
Every win is a template. Next county, next service, next question — the machine stamps citations while classic SEO is still waiting for month three.
gear five — where the machine stops being a trick and becomes a moat
✦
VII · old way vs new way
Getting recommended, two ways.
Old way
3–6 months · $200+/article
Pay a writer $200 per article, wait a week per draft
Chase keyword tools everyone else is using
Publish, then wait months for rankings to maybe move
Compete with nine other blue links for leftover clicks
The AI Overview quotes a competitor the whole time
Result: generic content, slow rank, invisible in AI answers
New way
~2 min per page · <24h to cited
Find questions the AI already answers (the hunt takes minutes)
Opus 5 writes the citation page in 107.9 seconds flat
Publish + push the index the same hour
The AI Overview lifts YOUR sentence — the next day
Template it across every county and service you serve
Result: the AI says your name first, at $5/M tokens
Thinking it?"Doesn't running the Agent OS burn a fortune in tokens?"
No — that's the biggest myth about it. The everyday 90% runs on free local models on your own machine, and free APIs slot in for more.
For the frontier work, the Agent OS drives the subscriptions you already own — Opus 5 shipped straight into Claude Code today, on the Claude plan you already pay for. And this guide's own test says run content on LOW effort: same page, fewer tokens.
Inside the AI Profit Boardroom there are full token-optimisation tutorials on top — usage drops further.
✦
VIII · three beliefs to drop
What's actually holding you back.
Wrong: "SEO takes six months. That's just how it is."
Right: RANKED takes months. CITED can take a day — the receipts above are timestamped. The AI answer lane is still early enough that a clean page wins fast.
Wrong: "I need to wait and see if Opus 5 is any good."
Right: It shipped this morning and wrote a complete, schema-correct citation page in 107.9 seconds on the first try — embedded above. The wait-and-see crowd is who you'll outrank.
Wrong: "AI Overviews are killing SEO, so why bother."
Right: They're killing PASSIVE SEO. For anyone willing to write the answer the AI wants to quote, they're the biggest shortcut search has offered in a decade.
Don't take my word for it
Members post their wins every day — agency owners, ecom founders, course creators, solo operators across 38 countries. Real businesses, real numbers, in their own words.
The Collin page is live, the search is public, and the workflow runs inside the same Agent OS 4,000+ founders use daily. Run the search yourself — it's linked in the sources panel.
Classic SEO waits for the crawler. The machine hands it the answer.
X · the SOP
Run the machine this week.
1
List 10 "best [your service] in [place]" searches. Your county, neighbouring counties, the cities you actually serve.
2
Run each one on Google. Keep the ones showing an AI Overview with no dominant cited source. Those are open citations.
3
Open Claude Code and switch to Opus 5. It's on your existing subscription — model claude-opus-5, effort LOW for content.
4
Prompt the citation page. Direct answer first, entity definition, service bullets, honest proof, GEO-vs-SEO table, prices, 5 FAQs with FAQPage schema, natural local mentions.
5
Read it before you publish. Fix anything you wouldn't say to a client's face. Honesty is the ranking factor.
6
Publish on your real domain. A page path like /best-[service]-[place] keeps the query match obvious.
7
Push it to the index the same hour. Search Console URL inspection → request indexing (or an indexing service). Never wait passively.
8
Check the search in 24–48 hours. Screenshot the citation when it lands — then feed the next county into the machine.
Thinking it?"My market's too competitive for this."
The hunt step filters for exactly that — you only build pages where the AI Overview exists but no strong source owns the citation yet. In local + service niches, that's most of the map right now.
Competitive head terms come later, funded by the easy citations you stack first.
✦
XI · the 30-day roadmap
From invisible to the recommendation.
Week 1
First citation
One query, one Opus 5 page, published + indexed. Watch the 24-hour clock yourself.
Week 2
The county sweep
Five neighbouring areas, same template. Low effort, ~2 minutes each.
Week 3
The service sweep
Every service you sell gets its own citation page. The map fills in.
Week 4
The moat
Refresh winners, screenshot citations, template the machine for clients. Now it's an asset.
Stop competing for the list. Start owning the sentence.
the recap
What you just gained.
i.
You stopped being invisible.
The target moved from "rank somewhere" to "be the sentence the AI says" — and you know how to win it.
ii.
You got the timestamped proof.
Published July 15, cited #1 in under 24 hours — real search, real page, both one click away.
iii.
You got a launch-day engine.
Opus 5, on your existing Claude plan, writes a complete citation page in 107.9 seconds.
iv.
You got the effort-toggle cheat.
Low effort = same content page, cheaper tokens. Save high for audits. Measured, not guessed.
v.
You got the template.
The exact page anatomy the AI lifts: direct answer, entities, FAQ schema, local proof.
vi.
You got the multiplier.
One winning page becomes a county map of citations at ~2 minutes per page.
