Checklist / Measurement Framework

A practical checklist for investigating search traffic loss

A falling traffic line does not identify its own cause. This checklist helps you collect comparable evidence, separate search surfaces and choose a specific investigation before changing content.

v1.1 · October 2026 · Tenten GEO Research

CH.01

Start with possible explanations, not a predetermined cause

Traffic can change because of measurement, access, demand, competition or search presentation. Several factors can overlap. A core-update date or an AI feature appearing near the drop is a lead to investigate, not a diagnosis.

The official May 2026 core update ran from May 21 to June 2, in U.S. Pacific time. Google’s May 19 announcement named Gemini 3.5 Flash as the new default model inside AI Mode and introduced a new Search box. It did not say that every Google search had become AI Mode by default. Keep an underlying model change distinct from a change in the search interface.

Check rankings and AI citations separately. Ahrefs’ March 2, 2026 study analyzed 863K keyword SERPs and 4M AIO URLs. Among already-cited URLs, 37.9% also appeared in the first ten SERP blocks and 37.1% in the top ten organic results. This does not give the probability that a top-ten page is cited or identify the cause of your site’s traffic loss. Assess content quality, answer passages and machine readability alongside your own query and traffic records.

Area	Observation	Next evidence	Reference
Measurement and access	Analytics or search traffic drops	Check data collection, availability, indexing and recent site changes	Google traffic-drop guidance
Search results and demand	Clicks, impressions or positions change	Compare the same queries and segments; inspect search presentation and demand	Search Console and Google Trends
AI citations	A target URL is absent from a captured answer	Record the engine, interface, prompt, answer date and linked sources	Scoped answer records; Ahrefs overlap study

Investigation areas, not a complete list of causes or automatic fixes.

Abstract pink and violet light lines crossing on a dark background — Conceptual artwork; the lines do not represent measured traffic or identified causes.

Chapter takeaway

First establish that the drop is real and define its scope. Then compare possible explanations against the evidence you can actually obtain.

CH.02

Use the right comparison window for a core update

For core-update analysis, Google recommends waiting at least one full week after rollout finishes before comparing Search Console data with a week before the update. This is timing guidance, not a randomized control or a reason to ignore an outage.

Use the actual start and completion dates in the Search Status Dashboard. Do not assume every update lasts the same number of weeks. Keep a dated record of site changes, incidents and collection changes so they can be considered during analysis.

A before-and-after window is a comparison, not a clean control group. Check whether the periods differ in seasonality, query mix, devices, countries or site availability. A longer context can help distinguish a recurring pattern from a new change.

STEP 1

Check urgent failures first

Confirm measurement and site availability. Repair a verified outage, broken access rule or accidental indexing block promptly, recording what changed; do not postpone it merely because a core update is running.

STEP 2

Record the actual update window

For the core-update comparison, wait at least a full week after the recorded completion date, following Google’s guidance. Preserve observations and avoid speculative mass edits.

STEP 3

Compare like with like

Compare affected pages and queries under consistent filters. Keep concurrent changes and unresolved differences visible; a stable difference still does not identify its cause by itself.

Chapter takeaway

Choose the comparison window carefully and keep the change log. Timing guidance improves interpretation; it does not make every remaining difference causal.

CH.03

Investigate ranking changes without a fixed fingerprint

A group of pages declining during an update deserves investigation. A single page can also change, and a site-wide drop can have a technical or measurement cause. No four-item checklist can prove or exclude a core-update effect.

Start with the affected queries and pages, the size and duration of the change, and the site’s own change history. A competitor moving above you is an observation about relative results, not proof of which ranking system changed or which edit will reverse it.

CHECK 1

Compare timing with the official record

An overlap makes the update worth investigating. A mismatch reduces that specific timing explanation but does not rule out all ranking changes or delayed observations.

CHECK 2

Locate the affected scope

Separate site-wide, topic, page and query changes. Use the pattern to choose what to inspect; do not map each pattern to a guaranteed cause.

CHECK 3

Read position, impressions and clicks together

Segment the same queries and pages. An aggregate average can hide changes in the mix; unchanged average position does not prove that AI caused a click decline.

CHECK 4

Inspect competing results and your own page

Check whether competing pages answer the task better and whether your page remains accessible and useful. Correct demonstrated issues instead of assuming every change requires a content rewrite.

Observation	Useful next check	Cannot establish alone
Many pages decline together	Availability, indexing, measurement and shared content changes	That the core update is the cause
One page declines	Page access, query demand, competing results and recent edits	That a ranking update is excluded
Dates overlap an update	Comparable before/after segments and concurrent changes	That timing proves causation
Competitor positions improve	Task relevance and the actual results shown	That one specific E-E-A-T or technical edit will restore the position

Use patterns to choose evidence, not to assign a diagnosis automatically.

Chapter takeaway

Record the observation, competing explanations and the next check. Keep the conclusion open when the available data cannot separate them.

CH.04

Distinguish AI Overviews, AI Mode and click changes

Seer’s AIO-prone query cohort had monthly organic CTR of 1.76% in June 2024 and 0.61% in September 2025. Its separately reported 61% decline has no explained calculation basis for those endpoints. BrightEdge separately reports AIO presence on roughly 48% of tracked keywords in February 2026.

AI Overviews and AI Mode are different Google Search experiences. An AI Overview appearing for a query does not show that the visitor used AI Mode. Lower CTR with stable impressions can also involve query mix, competing features, titles or snippets; inspect the relevant results and your own segments before assigning the change to AI.

Question	Evidence to collect	Interpretation limit
Was an AI Overview shown?	Dated observations for the affected query, locale and device	Does not identify AI Mode usage or a visitor’s click decision
Did clicks and impressions diverge?	The same query/page segments with both counts and CTR	Several presentation or mix changes can produce this pattern
Did positions change within segments?	Position distributions where available, not only a site average	A stable aggregate can hide different underlying changes
Which intent groups changed?	Comparable informational, comparison and transactional groups	A group difference does not prove that an answer displaced a visit

Identify the actual surface and denominator before making an attribution claim.

Chapter takeaway

Keep historical CTR change separate from AIO coverage. Neither study establishes the cause of an individual site’s traffic loss; compare your own rankings, impressions and clicks before choosing a fix.

CH.05

Read citation gaps as observations

In Ahrefs’ March 2, 2026 study, 37.9% of AIO-cited URLs also appeared in the first ten SERP blocks and 37.1% in the top ten organic results. This is the distribution of already-cited URLs, not a top-ten page’s citation probability or a fixed-cohort decline.

Ranking, citation and clicks are separate observations. Compare the same query, engine, date, ranking, cited URLs and click data. Absence from an answer does not prove that a passage was unreadable or that AI caused a traffic decline.

Chart 5.1

AIO-cited URLs that also appear in top-ten results

Ahrefs · 2026-03-02

Cited URLs: first ten SERP blocks

37.9%

Cited URLs: top ten organic results

37.1%

The denominator is already-cited AIO URLs. Ahrefs analyzed 863K keyword SERPs and 4M AIO URLs: for the same query, 37.9% appeared in the first ten SERP blocks, including ads and other features, and 37.1% in the top ten organic links. Improved parsing and different samples do not establish a decline for the same pages or diagnose an individual site. The unmatched original 17% source is retained in the research record and removed from this chart. Cite Tenten GEO and Ahrefs.

Question	How to inspect	What it can tell you
What is the organic position?	Record the same query, page, locale, device and date	One ranking observation; it does not rule out other ranking changes
Does the answer cite the target URL?	Retain the engine, interface, prompt and complete answer	Presence or absence in this sample, not a cause
Which sources appear instead?	Inspect the linked sources and their relevance to the question	Possible information differences to investigate, not proof of a readability defect
Can the intended system access the page?	Check actual responses, rendered text, indexing and the relevant provider’s access rules	A verified access issue can be repaired; allowed access still does not guarantee a citation

Keep ranking, citation and access evidence distinct.

Chapter takeaway

Track rankings, AI citations and clicks separately. If a top-ten page is absent from an AI answer, investigate crawl access, indexability, relevance and passage clarity. These are investigation paths; absence alone does not prove machine-readability failure.

CH.06

Choose the next investigation from the evidence

The checklist narrows what to inspect. It cannot identify a unique cause from three dashboard signals or promise recovery after one edit. Keep competing explanations and the evidence still missing in the report.

Use the table to choose a specific check and a result that would support or weaken the hypothesis. A useful outcome can be a confirmed defect, a narrowed explanation or a clear statement that the available data is inconclusive.

Observation	First check	Other explanations	Evidence needed	Action boundary
Traffic drops abruptly	Measurement and availability	Collection changes, access failures, indexing or demand	Logs, collection history and affected page samples	Fix verified failures and retain the change record
Positions decline	Affected queries and update dates	Competition, relevance, site changes and technical issues	Comparable segments and inspected results	Avoid broad rewrites based only on timing
Clicks decline while aggregate position holds	Query mix and search presentation	Demand, snippets, competing results or AI features	Matched query/page observations and click counts	Do not label all such changes AI Mode
A top-ten page is absent from an AI answer	The actual prompt, surface and cited sources	Sampling, relevance, access and answer variation	Repeated scoped observations and access evidence	Repair demonstrated issues; absence alone is not a diagnosis

A practical investigation guide, not a causal classifier.

TICK 1

Confirm the data and scope

Check collection changes and identify affected pages, queries and segments. Keep missing data and anomalies visible.

TICK 2

Check urgent access problems

Verify availability and unintended access or indexing blocks. Resolve demonstrated failures without waiting for an unrelated update window.

TICK 3

Choose a comparable period

For suspected core-update effects, use the official completion date and at least the following full week; consider seasonality and concurrent changes.

TICK 4

Distinguish the search surfaces

Record AI Overviews, AI Mode and other engines separately. A result in one surface does not establish behavior in another.

TICK 5

Test a specific explanation

Choose the smallest change that addresses a verified issue, record it and observe the relevant outcome. If the evidence remains inconclusive, say so.

A useful diagnosis states what remains unknown

The goal is a justified next action. Show what changed, what supports the explanation and what would change your conclusion. Do not turn uncertainty into a guaranteed cause or recovery promise.

Sources & methodology

Reviewed October 5, 2026. Google’s official dashboard records the May core update from May 21 to June 2, 2026 (U.S. Pacific time). https://status.search.google.com/incidents/wdAXJk6LRRihEjpzEeWE Its May 19 announcement describes a new default model inside AI Mode, not all searches becoming AI Mode. https://blog.google/products-and-platforms/products/search/search-io-2026/ Core-update comparison timing and traffic-drop investigation follow the official guides; the tables here are Tenten’s investigation framework, not a causal test. https://developers.google.com/search/docs/appearance/core-updates https://developers.google.com/search/docs/monitor-debug/debugging-search-traffic-drops AI Overviews and AI Mode are distinct surfaces. https://developers.google.com/search/docs/appearance/ai-features Ahrefs’ March 2, 2026 study covers 863K keyword SERPs and 4M already-cited AIO URLs: 37.9% overlap the first ten SERP blocks and 37.1% the top ten organic links. Different samples and parsing do not establish a fixed-cohort decline. Seer’s retrospective AIO-prone query cohort reports June 2024 and September 2025 monthly organic CTR of 1.76% and 0.61%, plus a separately reported 61% decline whose calculation basis is unresolved. BrightEdge’s roughly 48% is separate February 2026 tracked-keyword AIO coverage. These metrics cannot be combined into a shared bar scale, averaged or used to diagnose one site. https://www.seerinteractive.com/insights/aio-impact-on-google-ctr-september-2025-update https://www.brightedge.com/resources/weekly-ai-search-insights/ai-overviews-one-year-presence-size-citing https://ahrefs.com/blog/ai-overview-citations-top-10/

Discuss your AI search priorities

Request a 30-minute benchmark interpretation

Send a booking request. We’ll email you to confirm a time within 24 hours.