Track keyword ranking trends across years (not months)

The scenario

A keyword you used to rank #2 for has been hovering around #6 lately. You want to know exactly when it slipped, and whether it correlates with a Google update, an AI Overview rollout, or a competitor's content push. Standard GSC only shows 16 months. Manual export-and-pivot is brittle and tedious.

The prompt

Use the gsc-pap MCP. For https://example.com and the query 'customer interview templates', call historical_position_trend covering the last 24 months. Plot avg position by week (or by month if a finer granularity is too noisy). Annotate any week where position moved by 3+ points week-over-week. List those weeks with the position before and after, and call out any that align with known Google algorithm updates.

What happens

1. 1

   AI calls historical_position_trend for the query

   The tool returns a date-keyed series of (date, clicks, impressions, position) for the query. The longer the archive has been running on your account, the further back this goes, typically 18+ months for accounts authorised early.

2. 2

   AI smooths the noise and surfaces real moves

   Day-level ranking data is noisy (Google's serps shuffle hourly). The AI defaults to weekly aggregation, computes week-over-week deltas, and flags any week where position moves > 3 places, those are the weeks worth investigating.

3. 3

   AI annotates with context

   If you provide a list of known algorithm update dates (or ask the AI to look them up), it overlays them on the position chart and tells you which moves coincide with which updates. AI Overview rollout dates are a common 'aha', many sites lost rankings in late 2024 / early 2025 not because of quality drops but because the AI Overview ate the click slot above them.

Outcome

You stop guessing at why a keyword moved. Either you find a clear correlation with a known event (algorithm update, your own redesign, AI Overview), and act on it, or you confirm it's normal volatility and stop worrying. Either way: the data ends the speculation in 5 minutes.

Where to take it next

• ·Repeat for the top 20 queries to build a portfolio-level long-trend view.
• ·Pair with `compare_periods` for the same window if you also want clicks/impressions deltas.
• ·If the archive doesn't go far enough back yet, the dashboard at /account/archive lets you backfill the last 16 months Google still has on the live API.