In its October 15 Search announcement, Google said links embedded in AI Overviews and AI Mode responses would appear higher more often. It separately announced a test that highlights links to official pages for certain tasks. Publishers can check whether the new placements make their links more visible, but the announcement gives no measured increase in visits.
Jump to a section
Measure visibility and visits separately
Track the same pages and query types before and after the change. For each observation, record the date, query and result type, whether a link to the page appears, and where it appears. Separately, compare visits to the same pages across equal periods. The AI citation reporting guide shows why you should keep the original group of pages even when some disappear from a later report.
Google’s documentation on AI features explains that appearances in its AI features are included in overall Search Console Web reporting. That aggregate can show a change in search traffic, but does not by itself isolate visits from a link placed higher in an AI answer. If the placement is visible in a sample and visits rise, report those as separate observations unless you have evidence that connects them.
Separate more AI answers from more appearances in those answers
Keep a fixed list of queries. Record missed checks separately from completed checks that returned no AI answer. A rise in citations could mean more answers appeared, your pages appeared in more of those answers, or your team checked a different group of queries. A fixed panel lets you distinguish the first two explanations.
Imagine checking the same 50 queries in each of two weeks. In the first week, 20 checks produce an AI answer and eight answers contain your link. In the second week, 40 produce an answer and 12 contain your link. Your link appeared in more checks, rising from 16% to 24%. It appeared in a smaller share of AI answers, falling from 40% to 30%. Neither rate tells you whether the link moved higher.
For position, record whether the first link to your page appears inside the answer and which paragraph contains it. Use the same rule each week. If only three inline appearances qualify, list the observations. An average over three appearances can hide the difference between one prominent link and a broad shift.
| Measure | Week 1 | Week 2 |
|---|---|---|
| Checks with an AI answer | 20 | 40 |
| Answers linking to your site | 8 | 12 |
| Share of all checks with your link | 16% | 24% |
| Share of AI answers with your link | 40% | 30% |
How a rising total can mislead
Imagine 100 checks of the same queries in each period. Your pages appear in 20 answers before the change and 30 afterward. Meanwhile, total Google Search visits to those pages fall from 500 to 450. The sample’s appearance rate rose from 20% to 30%, while visits fell 10%. Both statements can be true. The example establishes no causal relationship between them.
A publisher can use that pattern to decide what to investigate. Check whether the rise is concentrated in a few queries, whether search demand changed, and whether the linked pages still serve the same reader intent. The citation definition distinguishes a source reference from the reader actions you can measure afterward.
What to change on your own site
Check that search can access the page and that the page answers the question behind the citation. The Google AI search requirements guide covers those checks. Keep the rollout and any site edits in a dated log so a later comparison can account for both.
Record the search surface you observed, too. A link appearing inside the answer and a link appearing beside it are different observations. Decide in advance whether your test cares about either placement or only the one Google said it would move.
Make the observation reproducible
For each check, record the date, device, location, query and page you expected to find. Mark whether the result was an AI Overview or AI Mode, whether it included your page and whether the link appeared inside the answer or alongside it. Save screenshots for a subset of checks so another person can classify them using the same rules.
If a colleague classifies one of those links as an inline citation and you classify it as a source card, resolve that difference before comparing weeks. Use the same rule on both sets of observations. If you change the rule, go back and recode the earlier sample.
What to report when you have no baseline
If you did not save observations before the announcement, begin with a dated snapshot. It can establish where your links appear in the current sample. Treat that as the baseline for later checks. There is no earlier observation to use for a before-and-after claim.
Set your observation dates beside the matching window for search performance. Label Search Console clicks and analytics visits separately because they measure different events. If links appear higher in later samples while visits stay flat, share both observations with their source and scope. A redesign of your content or a budget decision needs stronger evidence than the placement observation alone.
Measure citations against the number of sampled AI answers. AI answer citation rate calculator