LEARN

Measure the Baseline Before Touching Anything

Immobile basal matrix of fixed particles with overlying dynamic flow carrying shifting particles, embodying measuring baseline before action to attribute subsequent changes
Without a baseline, all subsequent changes cannot be attributed—unclear if caused by one's actions or market movement

IN ONE SENTENCE

Without a baseline, no later change can be attributed — you cannot separate your work from market drift.

A baseline is the starting measurement taken before optimisation. Without it, later changes cannot be separated from market drift — and there is no way to tell which investments worked.

OUR POSITION

Measure it before starting, on the same definition you will keep using. A baseline reconstructed afterwards is meaningless: content has changed and the crawl layer has been fixed, so what you measure is an intermediate state.

01

Which layers a baseline covers

Crawl: request volume, status codes and response sizes per crawler from server logs. Most often skipped, and it determines whether later explanations hold.

Index: indexable page count, actually indexed count, and the number of same-intent duplicates.

Answer: mention rate, average position and cited-source composition across a fixed question set, split by language.

Conversion: AI-channel visits and enquiries, plus current branded search volume and direct traffic.

02

Why third-party estimates cannot be the baseline

Third-party traffic and difficulty figures are estimates with opaque definitions that shift between tool versions. Using them for acceptance hands the interpretation to the vendor.

The correct use is prioritisation and competitive comparison, with acceptance on first-party analytics and server logs.

03

The step most often skipped

Archive the baseline together with its definition: the question set verbatim, the platform list, sampling runs, judgement rules and measurement date. Six months later, that archive decides whether the comparison holds.

⚠️ If the question set changes later, log the change too — which questions, when, and what the previous definition was.

Data behind this page

13.1 / 3.4 / 0.1 / ≈0

Monthly clicks per keyword at positions 1–3 / 4–10 / 11–20 / 21+

Source：Our keyword-level analysis of 4,074 non-branded keywords for a leading site in this category，2026-06

1.08%

AI referrals as a share of total site traffic

Source：Conductor, across 13,770 domains，2025

Sources

  1. [1]Optimizing your website for generative AI features on Google Search．Google Search Central．2026-05-15

Updated 2026-08-10