Engagement 02 — Recurring
A loop, not a retainer.
Monitor, find, improve, measure, repeat. Visibility moves, competitors move, models change and your own releases break things. A baseline from six months ago tells you nothing on its own — which is the honest argument for why this work is ongoing.
The loop
What actually happens each month.
Monitor
- AI mentions and citations across ChatGPT, Gemini and Perplexity
- Google organic rankings and traffic
- Google AI experiences where observable
- Competitor visibility on the same query set
- Backlinks and referring domains
- Reddit and community discussion
- Media coverage and brand mentions
- Review platform activity
Find
- New customer questions entering the market
- New query opportunities worth tracking
- Content gaps against the tracked set
- Competitor movement, separated from measurement noise
- Citation opportunities in the evidence chain
- Authority and PR openings
- Technical breakage introduced by your own releases
- Emerging topics before they are saturated
Improve
- Content: new pieces, refreshes, consolidation, pruning
- Technical SEO and AI crawler accessibility fixes
- Internal linking
- Structured data
- Entity consistency across external profiles
- Authority building
- Digital PR execution, human approved
- Expert-informed content from extraction sessions
Measure
- AI mention rate, with denominators
- Citation rate
- AI referral traffic where identifiable
- Organic traffic and rankings
- Leads and conversions
- Referring domains and brand mentions
- Revenue where attribution honestly allows it
Repeat
Measure, discover, prioritise, execute, measure again — against the same query set, with the same method. The compounding comes from two places: the work itself, and the history. By month six you can tell the difference between something you did and something the model changed, which is not a distinction anyone can make from a single snapshot.
Measurement discipline
Same queries. Same method. Every period.
Comparability is the whole point. If the query set changes, the comparison breaks — so query-set changes are recorded on the snapshot and flagged in the report rather than quietly absorbed.
- Queries are repeated within each period, so a rate is computed over executions rather than over single runs.
- Surfaces are reported separately. Averaging ChatGPT, Gemini and Perplexity into one number hides the thing worth knowing.
- Movement inside the noise band is reported as noise. We would rather tell you nothing happened than manufacture a win.
- Shipped work is timestamped on the timeline, so a later movement can be connected to something specific instead of claimed retrospectively.
Six periods against a fixed 42-query set. The annotations in the report note that comparison and alternatives pages shipped in July and two industry placements went live in September — which is how you connect movement to work rather than to hope.
Working together
What the engagement involves.
- Monthly
- Report, updated dashboard, shipped work, refreshed backlog, named blockers.
- Weekly
- Analysis pass, opportunity triage, anything urgent surfaced immediately rather than saved for the report.
- Quarterly
- Strategy review at exec level: trends, competitor shifts, what we are changing and why.
- Scope varies by
- Queries monitored, surfaces tested, competitors tracked, content output, digital PR effort.
- From you
- SME time, approval turnaround, and a path to getting code deployed.
Expectations
What a realistic first six months looks like.
No promises about outcomes. What follows is the shape of the work, not a forecast of results.
Month 1
Baseline locked before anything ships. Accessibility blockers fixed, because they are usually the fastest real change available. Expect very little movement in the numbers and a lot of movement in the backlog.
Months 2–3
First substantive content live. Entity consistency cleaned up across external profiles. Authority work started, which has the longest lead time of anything we do. Early movement, if any, will be inside the noise band and we will say so.
Months 4–6
Enough periods to distinguish trend from noise. Authority placements start landing. This is the point where the history becomes genuinely useful and the monthly conversation shifts from activity to direction.
A fair question
What happens in a month where the numbers go down?
We report it, in the same place and the same format as a month where they go up. Then we do the work of finding out whether it was your release, a competitor’s campaign, a model update, a change in how a surface selects sources, or ordinary variance in a non-deterministic system.
Sometimes the honest answer is that we cannot tell yet, and the correct response is another period of data rather than a hastily invented explanation. An agency that always has a confident story for every movement is not measuring carefully enough to know when it does not.
The loop starts with a baseline.
Which means it starts with the audit. Everything after that is comparison against a number you can trust.