The problem
I'll be honest, this started with a slightly uncomfortable test. On 5 Oct 2026 I ran a signed-out check across Google AI Overview, ChatGPT, Perplexity and Copilot to see whether any of them would cite simonhawtin.com. Not one did. Zero citations across the board.
ChatGPT and Copilot did know who I was, but only through LinkedIn, Finextra and scraped profiles, and they confused me with other Simon Hawtins.
The odd thing is that all 5 pages on my site were already indexed. So indexing wasn't the issue. The engines just weren't choosing my site as a source, and they weren't linking it to me. That's a different problem, and I felt like it needed a different way of working.
The system
So I built a small team of AI agents to plan, ship and measure Answer Engine Optimisation (AEO) changes to my site every week. Here's how it works.
- A strategy agent researches the target queries, ranks the backlog of site changes and writes one change per order, each with success criteria and an effort estimate.
- A website agent makes the edits in Claude Cowork and deploys to Netlify through a safe-deploy check. That means a preview first, 200 checks on the pages and schema validation.
- A chief-of-staff agent coordinates the work and throttles usage.
- A weekly measurement routine runs every Monday. It uses fixed prompts across Google AI Overview, ChatGPT and Copilot, signed out in Incognito so the results aren't personalised. It also tracks Google UK organic ranks for 10 keywords and logs everything to a running spreadsheet, with an alert whenever something changes.
- I approve each change. The agents don't deploy without a human yes from me.
It's genuinely a bit strange watching it all work, but the human gate is the part I'd never give up.
What shipped
Every change went out as its own order. In date order:
- 29 Sep 2026: robots.txt and sitemap.xml.
- Late Sep 2026: three pillar pages, demand gen leadership, AI in GTM and ABM leadership.
- Late Sep 2026: the Answer Engine Readiness Scorecard, the checklist I use to score AEO readiness.
- 19 Oct 2026: a definition block, a 5-question FAQ and FAQPage schema added to the scorecard.
Results so far
Here's what the checks have logged so far, all run signed out.
| Date | What I checked | Result |
|---|---|---|
| 5 Oct 2026 | AI answers across all engines, signed out | 0 citations |
| 14 Oct 2026 | Google AI answers, unpersonalised | First Google citations, 4 in total: AI Mode on branded and AI-in-GTM prompts, plus the AI Overview on a UK demand-gen leaders prompt |
| 19 Oct 2026 | Google UK organic, signed out | #1 for "demand gen leadership b2b saas" |
| 19 Oct 2026 | Google UK organic, signed out | #1 for "abm leadership b2b saas" |
| 19 Oct 2026 | Google UK organic, signed out | #1 for "answer engine readiness scorecard" |
| 19 Oct 2026 | Google UK organic, signed out | #5 for "demand generation director b2b saas" |
| 19 Oct 2026 | Google AI Overview | Cites simonhawtin.com for the demand gen and ABM leadership queries |
Going from zero citations to the top of Google UK for three topic keywords in about three weeks was more than I expected.
What hasn't worked yet
I want to be upfront about this part, because it matters as much as the wins.
- ChatGPT and Copilot still don't cite the site.
- Role queries such as "b2b saas marketing director" are dominated by job boards like Indeed and LinkedIn.
- The scorecard ranks #1 on Google UK but isn't in the AI Overview yet.
Next up is entity consistency, links from LinkedIn to the site, and mentions on other sites. I'm not sure yet which of those will move ChatGPT and Copilot first, so I'll be keeping a close eye on the Monday logs.
Lessons for B2B SaaS marketing leaders
If you're a B2B SaaS marketing leader thinking about AEO, this is what I'd take from it.
- Measure signed out. Personalised results lie, and they'll tell you you're winning when you aren't.
- Indexed isn't the same as cited. All 5 of my pages were indexed on 5 Oct and they still earned zero citations.
- Win topic queries before role queries. My topic keywords moved within weeks, whilst role queries are still owned by job boards.
- One change per order, each with success criteria. That way you know which change actually made a difference.
- Always keep a human approval gate. The agents do a huge amount of the work, but nothing goes live without my yes.
FAQ
What is an agentic AEO system?
It's a small team of AI agents that plan, ship and measure Answer Engine Optimisation changes to a site. On mine, a strategy agent writes one change per order, a website agent makes the edits in Claude Cowork and deploys to Netlify, a chief-of-staff agent coordinates the work and throttles usage, and I approve every change before it goes live.
How do you measure AEO results?
Every Monday I run fixed prompts across Google AI Overview, ChatGPT and Copilot, signed out in Incognito so the results aren't personalised. I also track Google UK organic ranks for 10 keywords, log everything to a running spreadsheet and get an alert whenever something changes.
How long did it take to get the first AI citation?
Nine days from my baseline. A signed-out check on 5 Oct 2026 found 0 citations, and on 14 Oct 2026 I got the first unpersonalised Google citations, 4 in total across AI Mode and the AI Overview.
Can this approach work for a B2B SaaS company brand?
I think so, but I'd be honest that my results come from a personal site, not a company brand. The method isn't specific to me though: measure signed out, ship one change per order with success criteria, win topic queries before role queries and keep a human approval gate.
Where to go next
If this was useful, the Answer Engine Readiness Scorecard is the best place to start. It's the checklist I use to score AEO readiness. You can also see how I approach demand gen leadership, AI in GTM and ABM leadership for B2B SaaS.
I'm Simon Hawtin, a B2B SaaS marketing and GTM leader based in London. If you're working on something similar, I'd genuinely love to compare notes on LinkedIn.