---
title: "Are Ads Crowding Out Your AI Citations? Run This 90-Day Query Test"
description: "Track ads, citation counts, and your domain’s share across your own search queries. Use a 90-day log to decide where citation work deserves investment."
url: "https://groas.com/post/are-ads-crowding-out-your-ai-overview-ci"
image: "https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/d5285450-2111-4c15-801c-59980666321c.png"
published: "2026-10-20T05:37:30.030Z"
modified: "2026-10-20T05:37:30.098Z"
---

October 20, 2026 · 11 min read

# Are Ads Crowding Out Your AI Citations? Run This 90-Day Query Test

[Alexander PerelmanHead Of Product @ groas](https://groas.com/author/alexander-perelman)[LinkedIn](https://www.linkedin.com/in/alexander-433793253/)

![Tiny model market square where blue "Sponsored" billboards rise over stalls while a figure with a clipboard counts the open stalls that remain.](https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/d5285450-2111-4c15-801c-59980666321c.png)

In this article

1. [Question: where can buyers still find you in an AI answer?](#question-where-can-buyers-still-find-you-in-an-ai-answer)
2. [Setup: choose 40 queries tied to the buying journey](#setup-choose-40-queries-tied-to-the-buying-journey)
3. [Controls: make each weekly run comparable](#controls-make-each-weekly-run-comparable)
4. [Procedure: record ads and citations by surface](#procedure-record-ads-and-citations-by-surface)
5. [Cadence: wait for a pattern, not a dramatic screenshot](#cadence-wait-for-a-pattern-not-a-dramatic-screenshot)
6. [Readout: three signals worth acting on](#readout-three-signals-worth-acting-on)
7. [Decision rules: put effort where the log shows room](#decision-rules-put-effort-where-the-log-shows-room)
8. [Limits: what your 90-day log cannot prove](#limits-what-your-90-day-log-cannot-prove)

An ad appears inside the AI answer for your best buying query. Your page does not. Before you rewrite your content plan or raise your bids, you need to know whether that result is a pattern in your market or one odd search session.

I made the broader case in [_Ads Will Eat the AI Overview: Five Predictions for Organic Citations Through 2027_](https://groas.com/post/ads-will-eat-the-ai-overview-like-they-a). This is the second look: predictions are easy to nod along to and hard to spend against. Ads are starting to appear in AI Overviews and conversational answers, but an industry forecast cannot tell you whether your own revenue queries still have citation space worth pursuing.

Here is the test I would run. Build a fixed set of 40 queries, check them weekly for 12 consecutive weeks within a roughly 90-day window, and log what appears under consistent conditions. At the end, decide whether to invest in pages that can earn citations, paid placement, or basic discoverability. **Your market’s results, not the industry forecast, make that call.**

## Question: where can buyers still find you in an AI answer?

For the searches that matter to your business, **how often do paid units appear, how many organic sources remain, and how often is your domain among them?** That is the operational question. Do not substitute a chart of overall AI Overview growth for an answer.

Your existing reporting will not settle it on its own. Google Ads Liaison Ginny Marvin [confirmed that ads inside AI Overviews and AI Mode are grouped into general top-of-page Search metrics](https://searchengineland.com/ginny-marvin-clarifies-ai-max-ai-search-ads-and-what-advertisers-should-prioritize-after-gml-479838), without separate placement reporting. Search Console does not give you a clean split between an organic citation click and a standard search-result click. To see the layout a buyer encounters, inspect and record the results directly.

My expectation is that paid presence will be strongest on high-intent searches: those are the queries where advertisers have the clearest reason to compete for attention. I would also expect citation opportunities to vary by query type and by the domains an answer engine trusts. Those are hypotheses, not findings from your account. The log is how you find out whether they hold.

## Setup: choose 40 queries tied to the buying journey

Do not pull 40 random keywords from an SEO tool. One broad industry average can mix buyer searches with recipes and celebrity birthdays. In Peec AI’s analysis of 500,000 buying-intent searches, [AI Overviews appeared on 86.7% of commercial queries and 88.5% of product comparison queries](https://www.searchenginejournal.com/google-ai-overview-data-looks-different-for-commercial-queries/577350/). The useful lesson for this protocol is simple: **sample the queries behind your pipeline, not the internet at large.**

Divide your fixed list into three buckets:

1. **15 high-CPC commercial terms.** Choose bottom-funnel searches with purchase intent, such as “b2b payment processing software” or “enterprise SOC2 compliance platform.” Check historical CPC in your Google Ads account so the bucket reflects commercial value in your market.
2. **15 mid-intent evaluation queries.** Include vendor comparisons, “top alternatives to \[Market Leader],” and “best \[software category] for mid-market teams.” These are the searches where a buyer is narrowing a list, not just learning a definition.
3. **10 diagnostic informational queries.** Use problem-aware searches a prospective buyer might make before comparing vendors, such as “how to calculate payment gateway churn” or “SOC2 bridge letter requirements.”

Freeze the query wording once the test starts. If you swap a weak query for a stronger one in week 6, give the replacement its own row and start its history then. Otherwise, a change in your test set can masquerade as a change in the search results.

## Controls: make each weekly run comparable

A logged-in browser can reflect your search history. [Practitioners tracking AI citations](https://www.reddit.com/r/SEO/comments/1uzsfwz/how_are_everyone_tracking_handling_citations_in/) have raised the same practical problem: a personal search session is a poor baseline. You cannot remove every source of variation, but you can stop changing the conditions yourself.

Hold these four controls steady for all 12 runs:

- **Location:** Use the same regional connection in your primary buyer market each week. Do not compare a Chicago result with one served from Ashburn as if they were repeat observations of the same market.
- **Session:** Start with a clean, logged-out browser state each run, or use an isolated automated browser session. Do not let last week’s clicks become this week’s test condition.
- **Device:** Keep the viewport fixed. If you test desktop at 1920×1080 and mobile at 390×844, treat them as separate suites. An ad that fits in one view may change what a reader sees in the other.
- **Time:** Run the set at the same hour and on the same day each week, such as Tuesday at 10:00 a.m. Eastern. Auctions and layouts can change between weekday business hours and off-peak periods.

Record any broken control. If your connection changes location or a run happens a day late, mark it. **A messy row is usable when you can identify it; an unmarked exception can mislead the whole readout.**

## Procedure: record ads and citations by surface

For every Google query, note whether an AI Overview appears. If it does, inspect three possible ad positions: above the answer, below it, and inside the answer. [This description of Google’s AI Overview ad zones](https://www.coseom.com/ads-in-ai-overviews/) gives you a useful visual checklist. Count the organic citation cards shown, record their domains, and note whether your brand receives a direct citation, an unlinked mention, or no mention.

Google’s answer interfaces have changed [from SGE to AI Mode](https://groas.com/post/from-sge-to-ai-mode-a-dated-history-of-h), so label the surface you actually inspected. A Google Overview and a conversational answer are not interchangeable observations. Do not pool them into one citation average.

![Diagram showing ad positions above, below, and within an AI Overview alongside organic citation cards](https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/5a5d393f-4a0f-4418-a24f-77670c54c430.png)

Run the evaluation and informational queries through ChatGPT, Perplexity, and Gemini as separate suites. For each response, record any paid unit you see, the number of cited URLs, the cited domains, and your brand status. If the first response has no sponsored unit, do not assume the conversation stays ad-free: Similarweb reported that [ChatGPT ads appeared in 26% of US desktop Free and Go sessions by June 2026, with 66.3% of sponsored placements appearing on or after the second prompt](https://aisearch.similarweb.com/blog/chatgpt-ads-reshaping-paid-search/). If you test a follow-up prompt, make it the same follow-up every week and label it separately from the opening answer.

Those platform observations can reveal opportunities, but they cannot tell you what happened in Google’s ad auction. **Compare each engine with its own previous runs first.** Cross-platform differences belong in your notes, not in a single blended score.

### Use one row per query, surface, and run

Set up a spreadsheet with the following nine fields. Use the same definitions every week; otherwise, your neat-looking chart will measure your changing judgment.

| Field                      | What to enter                                              | Why it matters                                           |
| :------------------------- | :--------------------------------------------------------- | :------------------------------------------------------- |
| **Query & bucket**         | Exact query; Commercial, Evaluation, or Informational      | Keeps intent groups distinct                             |
| **Run date & week**        | Date; Week 1 through Week 12                               | Shows when a change appeared                             |
| **Engine & surface**       | Google AIO, ChatGPT, Perplexity, or Gemini                 | Prevents unlike answers from being pooled                |
| **Answer triggered**       | Yes or No                                                  | Separates absent answers from uncited ones               |
| **Ad zone detected**       | None, Pre-Answer, Post-Answer, In-Answer, or a combination | Locates paid presence                                    |
| **Organic citation count** | Number displayed                                           | Tracks available source slots                            |
| **Cited domains**          | Root domains shown                                         | Reveals recurring sources                                |
| **Brand status**           | Cited (Link), Mentioned (No Link), or Absent               | Distinguishes a link from a passing mention              |
| **Classic organic rank**   | Rank or Unranked; Google only                              | Lets you compare citation presence with standard results |

Count citations the same way each time. If a card expands to show more sources, decide before week 1 whether you will count only what is initially visible or the full expanded set. Keep that rule. Your brand’s _citation share_ can then be calculated as the share of eligible answer runs in which your domain received a linked citation. Keep mentions without links separate. They may be interesting; they are not citation wins.

## Cadence: wait for a pattern, not a dramatic screenshot

Run the complete set once a week for 12 consecutive weeks, then leave enough room in the roughly 90-day window to review exceptions and calculate the results. A single result can reflect a layout test or ordinary variation. Even a clean session does not guarantee you will see the same answer twice.

For each query, calculate how many runs showed an answer, an ad in each zone, and a linked citation for your domain. Then compare early and late weeks within the same query and surface. **Frequency matters more than the most alarming screenshot.** An in-answer ad seen twice in 12 runs deserves monitoring. One seen in 10 runs deserves a closer look at your paid and organic options. Neither figure proves that the placement is permanent.

Do the same with citation counts. If the count falls once and rebounds, note it. If it remains lower across several consecutive runs, investigate. The point of the cadence is not to declare a winner after week 2; it is to keep a volatile interface from setting your budget on a whim.

![Editorial illustration of a growing paid ad block narrowing the space around organic citation cards](https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/20ca7d8d-2ef7-48eb-a089-770b745a367a.png)

## Readout: three signals worth acting on

At week 12, sort the log by query bucket, then inspect these signals in order. Keep Google, each conversational engine, and each device suite separate.

1. **Paid presence by intent.** Compare your high-CPC commercial queries with your informational ones. If ads repeatedly appear on the former and rarely on the latter, paid inventory is clustering around purchase intent in your sample. If not, do not force the expected story onto the data. Check which exact queries account for the difference before changing bids.
2. **Citation count alongside ads.** Compare citation counts on the same queries over time, noting when ads were present. A smaller count on ad-heavy answers is a reason to examine available space, not proof that an ad caused a citation to disappear. Different queries can have different source needs; a comparison between buckets alone will not establish cause.
3. **Who holds the remaining slots.** List the domains cited repeatedly on your commercial and evaluation queries. If directories such as G2, Capterra, or Forbes Advisor occupy most of the visible slots while vendor domains seldom appear, a vendor page may face a different citation challenge from a directory. That pattern tells you where to focus your next content decision. It does not tell you why an engine chose each source.

Next, check your own domain’s linked citation share in each bucket. A page that ranks in classic results but never earns a citation presents a different problem from a domain that appears nowhere at all. **Make the decision at the query-bucket level, then inspect the individual queries before you spend.**

## Decision rules: put effort where the log shows room

The thresholds below are operating labels, not laws of search. Use them to choose a next move, then keep measuring whether that move helps.

1. **Open citation space: few ads, five or more citations.** If commercial or evaluation answers repeatedly show multiple vendor sources and little in-answer paid presence, prioritize pages an answer engine can extract cleanly. Put a direct answer near the top of a relevant solution page, make comparison tables readable in HTML, and remove copy that hides the actual specifications. Do not stop useful paid campaigns merely because an organic slot exists. Seer Interactive’s research [reported more organic and paid clicks for cited brands on overview queries](https://ppc.land/google-ai-overviews-reduce-organic-ctr-61-paid-traffic-68/); your test still needs to show whether your pages can enter those answers.
2. **Paid-heavy answers with few vendor citations.** If a commercial query repeatedly carries ads while its one to three citations are dominated by directories, do not spend months rewriting the same vendor page in the hope that another adjective will get it cited. Examine the paid opportunity on that query, the conversion value you send back to your bidding system, and whether your existing targeting reaches the right buyer. Keep the citation work that has a plausible path; move the rest of the effort toward placements you can test.
3. **No brand presence anywhere.** If your domain is neither cited nor represented in ads across the relevant runs, address discoverability before debating citation-share tactics. Start with the pages a buyer needs to evaluate you: a pricing or cost explanation, a candid comparison page, and a clear account of how the product works. Then rerun the same queries. If visibility remains absent, at least you have ruled out the most basic content gap.

## Limits: what your 90-day log cannot prove

This is a controlled market check, not a universal study. **Forty queries and 12 weekly runs can guide your next decision; they cannot establish how every buyer saw every result.** Keep these limits attached to any chart you share:

- **Answers vary.** The same prompt can return different citations minutes apart. Repeated observations help you see a pattern, but they do not turn an AI answer into a fixed ranking.
- **Markets vary.** A result in one location or on desktop may not match mobile or another country. Run separate suites if those audiences matter to your revenue.
- **Session and rollout effects remain.** Clean browser state reduces one source of bias; it does not remove search-engine experiments or regional changes. Look for sustained movement before overhauling pages or bidding.
- **Presence is not revenue.** An ad or citation in an answer does not, by itself, tell you whether a qualified buyer clicked or converted. Read the visibility log beside your existing pipeline and acquisition data before you reallocate a meaningful budget.

Repeat the suite quarterly. Keep the original 40 queries as your comparison set, and track any new queries separately. By day 180 or day 270, you should be able to see whether the conditions that drove your first decision persisted or changed. That is more useful than reopening the same industry prediction every time an AI interface gets a new coat of paint.

The result can go either way. If your queries still show accessible citation slots, build and measure pages that deserve them. If ads recur and vendor citations stay scarce, test the paid path instead of funding a content rewrite on faith. And if your domain is absent everywhere, fix that before arguing about whether ads are swallowing it.

The spreadsheet makes the choice visible; it does not execute it. Traditional agencies can charge a retainer to turn that choice into another deck. At [groas](https://groas.com/), we built an autonomous growth engine to manage paid and organic search continuously, with specialized AI models doing the execution and a named human strategist setting direction and guardrails. If your log shows what needs to change and you do not want another recommendation to implement yourself, [apply for a free trial](https://groas.com/for-businesses).

## Frequently Asked Questions

### Why should I run a custom test instead of relying on industry forecasts?

The test provides data on how ads appear, how many organic sources remain, and whether your domain is cited within your specific revenue queries. Industry forecasts do not accurately reflect whether your own revenue queries still have citation space worth pursuing.

### How should I select the 40 queries for the test?

Queries should be sampled from your sales pipeline, not the internet at large. The 40 queries must be divided into 15 high-CPC commercial terms, 15 mid-intent evaluation queries, and 10 diagnostic informational queries.

### What controls must I keep steady during each weekly test run?

You must maintain fixed conditions for location, session, device, and time across all 12 runs. This protocol prevents internal variations from skewing the results and ensures each weekly observation is comparable.

### What key data points should I record for every query run?

The log must note the AI Overview appearance, the ad zone detected (such as above, inside, or below the answer), the organic citation count, and your brand's specific status—cited, mentioned, or absent.

### What should I do if my queries show open citation space?

If commercial or evaluation answers consistently show multiple vendor sources and little paid presence, prioritize clean content. Put a direct answer near the top of a relevant solution page and ensure comparison tables are readable in HTML.

### What limitations should I keep in mind when reviewing the test results?

The 90-day log is a controlled market check, not a universal study. Results from one location or on one device may not match others. An ad or citation in an answer does not, by itself, confirm a buyer clicked or converted.

## Pay For Results, Not For Hours

Businesses buy the outcome, agencies resell it, and groas answers for it either way.

[See If You Qualify](https://groas.typeform.com/to/xC1bQNUT)

## Structured data

```json
{"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://graph.groas.com/entity/groas","name":"groas","url":"https://groas.com/","logo":{"@type":"ImageObject","url":"https://groas.com/icon-512.png","width":512,"height":512},"sameAs":["https://www.linkedin.com/company/groas/"]},{"@type":"WebSite","@id":"https://groas.com/#website","name":"groas","url":"https://groas.com/","inLanguage":"en-US","publisher":{"@id":"https://graph.groas.com/entity/groas"}},{"@type":"WebPage","@id":"https://groas.com/post/are-ads-crowding-out-your-ai-overview-ci#webpage","url":"https://groas.com/post/are-ads-crowding-out-your-ai-overview-ci","name":"Are Ads Crowding Out Your AI Citations? Run This 90-Day Query Test","description":"Track ads, citation counts, and your domain’s share across your own search queries. Use a 90-day log to decide where citation work deserves investment.","inLanguage":"en-US","isPartOf":{"@id":"https://groas.com/#website"},"about":{"@id":"https://graph.groas.com/entity/groas"},"primaryImageOfPage":{"@type":"ImageObject","url":"https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/d5285450-2111-4c15-801c-59980666321c.png"},"breadcrumb":{"@id":"https://groas.com/post/are-ads-crowding-out-your-ai-overview-ci#breadcrumb"},"dateModified":"2026-10-20T05:37:30.098Z","mainEntity":{"@id":"https://groas.com/post/are-ads-crowding-out-your-ai-overview-ci#article"}},{"@type":"BreadcrumbList","@id":"https://groas.com/post/are-ads-crowding-out-your-ai-overview-ci#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://groas.com/"},{"@type":"ListItem","position":2,"name":"Blog","item":"https://groas.com/blog"},{"@type":"ListItem","position":3,"name":"Are Ads Crowding Out Your AI Citations? Run This 90-Day Query Test","item":"https://groas.com/post/are-ads-crowding-out-your-ai-overview-ci"}]},{"@type":"BlogPosting","@id":"https://groas.com/post/are-ads-crowding-out-your-ai-overview-ci#article","url":"https://groas.com/post/are-ads-crowding-out-your-ai-overview-ci","mainEntityOfPage":{"@id":"https://groas.com/post/are-ads-crowding-out-your-ai-overview-ci#webpage"},"isPartOf":{"@id":"https://groas.com/blog#blog"},"headline":"Are Ads Crowding Out Your AI Citations? Run This 90-Day Query Test","description":"Track ads, citation counts, and your domain’s share across your own search queries. Use a 90-day log to decide where citation work deserves investment.","datePublished":"2026-10-20T05:37:30.030Z","dateModified":"2026-10-20T05:37:30.098Z","image":{"@type":"ImageObject","url":"https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/d5285450-2111-4c15-801c-59980666321c.png"},"author":{"@id":"https://groas.com/author/alexander-perelman#person"},"publisher":{"@id":"https://graph.groas.com/entity/groas"},"keywords":"citation, ads, queries, query, run, test, answer, each","wordCount":2410,"inLanguage":"en-US"},{"@type":"Person","@id":"https://groas.com/author/alexander-perelman#person","name":"Alexander Perelman","jobTitle":"Head Of Product @ groas","description":"Ex Goldman Sachs and Ex Stanford Computer Science","image":"https://groas.com/media/blog/4f17cac3a81acc224cc1b2cfab8f94bbc295a3089d13e8421125771ef8d4064f.jpg","worksFor":{"@id":"https://graph.groas.com/entity/groas"},"sameAs":["https://www.linkedin.com/in/alexander-433793253/"]},{"@type":"FAQPage","@id":"https://groas.com/post/are-ads-crowding-out-your-ai-overview-ci#faq","mainEntity":[{"@type":"Question","name":"Why should I run a custom test instead of relying on industry forecasts?","acceptedAnswer":{"@type":"Answer","text":"The test provides data on how ads appear, how many organic sources remain, and whether your domain is cited within your specific revenue queries. Industry forecasts do not accurately reflect whether your own revenue queries still have citation space worth pursuing."}},{"@type":"Question","name":"How should I select the 40 queries for the test?","acceptedAnswer":{"@type":"Answer","text":"Queries should be sampled from your sales pipeline, not the internet at large. The 40 queries must be divided into 15 high-CPC commercial terms, 15 mid-intent evaluation queries, and 10 diagnostic informational queries."}},{"@type":"Question","name":"What controls must I keep steady during each weekly test run?","acceptedAnswer":{"@type":"Answer","text":"You must maintain fixed conditions for location, session, device, and time across all 12 runs. This protocol prevents internal variations from skewing the results and ensures each weekly observation is comparable."}},{"@type":"Question","name":"What key data points should I record for every query run?","acceptedAnswer":{"@type":"Answer","text":"The log must note the AI Overview appearance, the ad zone detected (such as above, inside, or below the answer), the organic citation count, and your brand's specific status—cited, mentioned, or absent."}},{"@type":"Question","name":"What should I do if my queries show open citation space?","acceptedAnswer":{"@type":"Answer","text":"If commercial or evaluation answers consistently show multiple vendor sources and little paid presence, prioritize clean content. Put a direct answer near the top of a relevant solution page and ensure comparison tables are readable in HTML."}},{"@type":"Question","name":"What limitations should I keep in mind when reviewing the test results?","acceptedAnswer":{"@type":"Answer","text":"The 90-day log is a controlled market check, not a universal study. Results from one location or on one device may not match others. An ad or citation in an answer does not, by itself, confirm a buyer clicked or converted."}}]}]}
```