Most agency owners I talk to fall into two AI camps.
Camp 1: "We use ChatGPT sometimes for first drafts." (Translation: they barely use it.)
Camp 2: "We've automated our entire content workflow." (Translation: they're publishing AI slop at scale and quietly losing rankings.)
Neither extreme works.
The agencies winning right now use AI as a force multiplier on senior judgment — not as a replacement for it. The juniors get accelerated. The seniors get freed up. The output quality goes up, not down.
This article walks through the exact AI SEO workflow we run at Goldie Agency across 300+ client engagements. Each step lists the prompt(s), the tool, the human checkpoint, and the time saved.
If you run an agency or in-house SEO team, copy this directly.
The 7-stage workflow at a glance
Every client engagement at Goldie runs through these 7 stages. AI is layered into 5 of them. The other 2 stay 100% human.
- Discovery & audit (AI-augmented)
- Strategy & roadmap (100% human)
- Keyword + content planning (AI-augmented)
- Content production (AI-augmented)
- Link building & outreach (AI-augmented)
- Reporting & client comms (AI-augmented)
- Quarterly strategy review (100% human)
Let me walk through each.
Stage 1 — Discovery & audit
When a new client signs, you need a comprehensive audit fast. This used to take a senior 2-3 days. We now turn it around in 6 hours.
What AI does here
- Technical audit summarisation. Run Screaming Frog or Sitebulb, paste the raw issues list into ChatGPT with a prompt that ranks them by impact and groups them into themes.
- Competitor SERP audit. Drag SERP screenshots into ChatGPT (vision), ask it to identify the content patterns, SERP features, and intent signals Google is rewarding.
- Backlink gap analysis. Export the client's referring domains + top 3 competitors' referring domains, paste into Claude with a gap-finder prompt.
What stays human
- Reading the actual site. The senior strategist still spends 90 minutes inside the client's CMS, analytics, and core pages. AI doesn't catch UX intent issues.
- Talking to the client. The 60-minute discovery call is always human-to-human.
Time saved
Day 1: 2-3 senior days → 6 hours. About 75% faster, with equal or better depth.
Stage 2 — Strategy & roadmap (100% human)
This stays human because it's the highest-leverage decision in the whole engagement.
The senior strategist takes the audit outputs and writes a 90-day roadmap — what we'll prioritise, why, in what order, and what success looks like.
AI can draft summaries and pull data, but the actual decision about what to do and when is judgment work. It's where seniors earn their salary.
We don't shortcut it.
Time saved
Zero. By design.
Stage 3 — Keyword + content planning
This is where AI really starts pulling weight.
What AI does here
- Keyword cluster building. One prompt produces the entire cluster map for a pillar topic. Output: pillar keyword, 8-12 supporting keywords, intent classifications, internal linking suggestions.
- Content gap analysis. Paste competitor URLs, get back a prioritised list of topics they cover that we don't.
- Brief generation. Every article gets an AI-generated brief that includes the differentiation angle from the SERP. Senior strategist reviews and approves in 5 minutes.
- Editorial calendar drafting. AI proposes a 90-day editorial calendar based on the cluster map + cluster maturity model.
What stays human
- The strategist approves every brief before it goes to a writer. About 70% pass first review, 20% get notes, 10% get redirected.
- Final editorial calendar approval. AI proposes, senior decides.
Time saved
What used to be 2-3 days of strategist time per quarter is now 4-6 hours. About 80% faster.
Stage 4 — Content production
This is where most agencies are blowing it.
The "publish 100 AI-generated posts a month" pitch is a trap. Sites doing this in 2026 are getting hammered by Helpful Content updates and watching their rankings collapse.
We use AI to accelerate writers, not replace them.
What AI does here
- First draft generation. Writer takes the brief, runs a draft-generation prompt to get a structured first draft (Claude does this best). Then they rewrite in their voice, add original examples, and inject opinion.
- Internal linking suggestions. After draft is done, AI scans against the site's existing posts and proposes 5-8 internal linking opportunities.
- Schema + FAQ generation. AI generates the FAQ section and corresponding JSON-LD schema markup in one prompt.
- Title tag and meta description optimisation. AI generates 10 options per article. Writer or editor picks the strongest.
What stays human
- The voice. The opinion. The examples. The on-the-ground insight that makes content actually rank in 2026.
- Editorial review. Every article goes through a senior editor before publishing.
Time saved
A senior editor + writer team that used to ship 8 articles a month now ships 16-20, at the same or better quality.
Stage 5 — Link building & outreach
The single highest-leverage AI application in agency SEO.
What AI does here
- Prospect list generation. Export competitor backlinks, run them through an AI prioritisation prompt that scores each by likelihood of conversion + relevance + DR.
- Contact finding. Tools like Apollo + ChatGPT for fallback identification. (ChatGPT can often guess email formats from a domain pattern.)
- Personalisation at scale. Custom GPT trained on our outreach voice generates personalised first sentences for each target by reading their recent content.
- Follow-up sequencing. AI drafts the day-4 bump and day-10 pivot emails based on the original email + the target's profile.
- Reply triage. AI classifies inbound replies as "yes," "no," "ask for more info," or "pivot" — junior outreach manager only opens the ones worth attention.
What stays human
- The first 30 prospects on every list. The senior outreach lead reads them before any email goes out. Catches the targets that AI mis-prioritised.
- Negotiation. When a placement requires a real back-and-forth, a human takes over.
- Anything bespoke. Digital PR pitches to Forbes-tier publications are still 100% human.
Time saved
A campaign that used to take a senior outreach lead 80 hours of work now takes 25 hours of senior time + 30 hours of junior + AI time, for higher placement volumes and equivalent quality.
Stage 6 — Reporting & client comms
The least sexy but highest-volume time sink in any agency.
What AI does here
- Monthly client report drafting. Paste in GSC export, Ahrefs export, and placement list. AI drafts a 1,500-word executive narrative in the senior consultant's voice.
- Slack/email comms. AI drafts client-facing updates that the account manager edits in 30 seconds.
- QBR slide deck generation. AI builds the first draft of the quarterly business review deck — strategist edits the strategy slides, AI handles the data slides.
What stays human
- The "we screwed up this month" honesty conversation. Always.
- Strategic recommendation slides in the QBR. AI drafts the data; humans write the recommendations.
- Anything that goes to the client's CEO or board.
Time saved
Account management hours per client dropped from 12-15 per month to 4-6.
Stage 7 — Quarterly strategy review (100% human)
Every 90 days, the senior strategist sits with the client and reviews what's working, what's not, and what to change.
This stays 100% human. The AI doesn't get a seat at this table.
Time saved
Zero. By design.
The total picture
Across the 7 stages, AI saves us roughly 60% of the operational time we used to spend per client — at equal or higher quality.
That 60% gets reinvested in two places:
- More clients per senior (capacity expansion).
- Deeper work on each client (quality expansion).
We chose option 2. That's what's let us grow Goldie Agency revenue per client by 40% over the last 18 months without hiring proportionally.
If you're an agency owner, the math is straightforward: AI doesn't make your agency cheaper. It makes the seniors you already have meaningfully more effective.
Where to start if you're behind
If you're not running this workflow yet, here's the order I'd add stages:
- Start with Stage 6 (reporting). Lowest risk, highest weekly time saving. Get this working first.
- Then Stage 5 (outreach). Highest revenue impact.
- Then Stage 3 (briefs). Quality multiplier.
- Then Stages 1 and 4. These need the most senior oversight, so add them once the team has confidence.
What's next
Every prompt referenced in this workflow is in the free AI SEO prompt library.
If you'd rather have me run this workflow on your account directly, book a strategy session with Goldie Agency. Free, 30 minutes, with me.
If you want to compare notes with other senior operators running variants of this workflow, that's what the SEO Elite Circle is for.