How do you choose an AI search agency?
Choose an AI search agency by how it measures, not by what it promises: it should ask each question several times on the real consumer apps, report a range for each app, and prove change against a control group of questions it never touched. Then ask for published prices, a written list of the tactics it refuses, and the name of the person who does the work. KailxLabs publishes all of this, starting with a $1,500 AI visibility baseline.
What should the checklist cover?
Eight things separate an agency that can prove its work from one that sends screenshots.
| Check | A good answer | A red flag |
|---|---|---|
| How it measures | Named consumer apps, each reported on its own | One blended "AI visibility score" |
| Apps or APIs | The real apps, such as ChatGPT on chatgpt.com; any API data labelled | API answers presented as what buyers see |
| Repeated runs | Several runs per question, shown as a count and a likely range | One answer per question, or screenshots |
| Control group | Part of the questions held back untouched, so change can be proven | Before and after charts with nothing to compare against |
| Published prices | Prices and terms on the website | Prices only after three calls |
| Refused tactics | A written list: no fake reviews, no planted threads, no invented numbers | Review campaigns, seeded Reddit posts, self ranked lists at volume |
| Who does the work | A named senior person approves every change | Unnamed juniors and a content quota |
| Promises | A clear guarantee tied to measured answers, or none | A guaranteed rank or position in ChatGPT |
Why measurement comes first
Answers change from run to run and from app to app. An agency that measures badly cannot tell you whether its work did anything.
The ChatGPT API and app shared only 12.0% of cited domains on the same questions.
arXiv 2609.18729, academic audit · 1,536 answers · 16 Sep 2026
Five runs per prompt is exploratory; 10 is the level for confirming a claim.
Dice Roll Method, arXiv 2609.04047 · Reanalysis of about 190,000 observations · 3 Sep 2026
The IAB asks for repeated runs, per platform results and ranges.
IAB, "Measuring Visibility in the AI Era" · Industry framework · 3 Aug 2026
Questions to ask any agency
Ask all eight. Good agencies answer in writing without hesitating.
- Which AI apps will you measure, and do you read the consumer app or an API?
- How many times do you ask each question, and how do you show the range?
- Which questions will you hold back as a control group?
- What counts as being named: an exact brand match, a link, or a recommendation?
- Can I see the raw answers behind every number in your report?
- Which tactics do you refuse, in writing?
- Who will do the work on my account, and who approves it?
- What happens if nothing moves after 90 days, and can I leave?
Why agency rankings mislead
Most "best AEO agency" lists are written by an agency that ranks itself. They tell you who publishes lists, not who does good work.
91 of 108 agency lists we parsed were published by an agency ranking itself, usually first. Only 7 came from independent publishers.
KailxLabs review of agency listicles · 108 lists from 50 search result pages · 26 Sep 2026
Some directories sort sponsored firms first by default, and some list placements are sold as a service. Ask whether any ranking you read was paid.
Directory and agency pages reviewed · 26 Sep 2026
Check us the same way
AI visibility baseline
$1,500 one time. The baseline fee is credited in full if you start a program within 30 days.
About the baselineHow we measure
Consumer apps named, repeated runs, ranges, and an untouched half of questions held back for proof.
MethodologyPublished prices
Programs from $2,000 a month. Programs run month to month after the first 90 days.
PricingChoosing an agency, answered.
Should I pick an agency from a "best AEO agencies" list?
Use those lists with care. In our review of 108 such lists on 26 Sep 2026, 91 were published by an agency ranking itself, usually first. Judge agencies on method and terms instead.
How many questions should an agency measure?
At least 50. The IAB's August 2026 framework calls fewer than 50 queries exploratory, and wants all four intent types covered: research, comparison, recommendation and buying.
Is a long contract a red flag?
It is a question worth asking. AI answers change quickly, and you should be able to leave if the method is not working. At KailxLabs, programs run month to month after the first 90 days.
Does KailxLabs pass its own checklist?
We built it from our method. Prices are published, measurement runs on the consumer apps with repeated runs and a control group, and the founder approves every change. Nobody controls what an AI says. We never promise a fixed spot or a rank.
- 26 Sep 2026: page published.
Talk to the founder.
20 minutes with Kailesk Khumar. You leave knowing where AI names you today, and what it would take to become the answer.
- A live look. We ask five of your buyer questions on ChatGPT, Google and Perplexity while you watch.
- The reasons. The three biggest reasons AI names a rival instead of you.
- The plan. A clear next step and its price, whether or not you hire us.
Pick a time that suits you. The calendar invite arrives straight away. Prefer a link? Book on Cal.com.