AEO tools are software that asks AI engines such as ChatGPT, Gemini, Perplexity and Claude the questions your buyers ask, then records whether your brand is named, how it is described and which pages are cited. To choose one, check engine coverage, prompt volume, citation data, reporting and attribution, then test it on your own brand before paying.
Key takeaways
- Pick on five criteria: engine coverage, prompt volume, citation data, attribution and reporting. Price comes after fit.
- Any tool that shows a single "ranking position" in AI deserves suspicion. Repeat-run research says mention rate across many runs is the steadier number.
- Ask for the stored answer behind every number. If you cannot read the raw answer, you cannot audit the metric.
- Use the free trial as a test, not a tour: load your own prompts, check the answers by hand, and use each of the three days for a set job.
- Attribution is a separate layer. Most AEO tools measure the answer; revenue still needs referrals, analytics and a "how did you hear about us" field.
- Start with a free baseline from the Noma by Nurdd free brand audit so you know what a paid tool should improve on.
What are AEO tools?
AEO (answer engine optimisation) tools monitor and measure how a brand appears inside AI-generated answers. AirOps, 2026 describes the category as tools that help marketing teams monitor, influence and measure brand appearance in AI search, tracking metrics such as citation rate, mention rate, sentiment and competitive positioning. Some also audit your site or draft content.
Three kinds exist. Read-only trackers record answers. Audit tools score your pages for machine readability. Workflow tools connect findings to content briefs and drafts. Most buyers need a tracker first, because you cannot fix what you have not measured. For the wider category and a vendor roundup, see Best AI Visibility and AEO Tools in 2026; this guide covers how to choose, not who is on the list.
Why you need to test AEO tools before buying
AI answers vary run to run, so a tool can look precise while measuring noise. Two public studies show how much.
SparkToro and Gumshoe.ai ran 2,961 prompts across ChatGPT, Claude and Google's AI search features. According to Search Engine Journal, January 2026, the chance of ChatGPT or Google's AI returning the same brand list across repeated runs of one prompt was under 1 in 100. Visibility across many runs was steadier than position.
"If you ask an AI tool for brand/product recommendations a hundred times nearly every response will be unique." Rand Fishkin, SparkToro co-founder, as quoted by Search Engine Journal, January 2026.
Citations move too. Ahrefs followed about 37,500 citations across five AI tools daily for 20 days. Ahrefs, October 2026 reports that 50 to 83% of citations were gone the next day, and about 75% of the lost citations returned during the 20-day study. Its advice: do not judge visibility from one check.
The practical consequence: a tool that runs each prompt once, or reports a precise position from a single answer, gives you a number you cannot defend to a CFO. Your selection criteria should reward repetition, stored answers and stated denominators. AI Citation Stability explains how many runs to sample.
How do you choose an AEO tool? The five-criteria checklist
Score each tool from 0 to 2 on each row. A tool that scores 0 on stored answers or engine coverage for your market should be dropped whatever its price.
| Criterion | What to ask the vendor | Pass looks like | Red flag |
|---|---|---|---|
| Engine coverage | Which engines, in which versions, and which are included in my plan? | The engines your buyers use, listed by name, with the interface or API stated | "All major AI engines" with no list |
| Prompt volume | How many prompts, how often are they run, and are runs repeated? | A fixed prompt set, run on a schedule, with the limit stated per plan | One run per prompt, or limits hidden until checkout |
| Citation data | Can I see the URLs and domains each answer cited, and the raw answer? | Cited URLs per answer, grouped by domain, with the stored text | Only an aggregate score |
| Attribution | What links answers to visits and pipeline? | Referral and crawler data, plus an export or API to your analytics | A revenue number with no method |
| Reporting | Can I filter by engine, prompt group and date, and share a report? | Per-engine trends, competitor share of voice, scheduled reports | One blended dashboard score |
Engine coverage
Engines disagree, and the gap is large inside one company's products. An SEJ report, January 2026 notes Ahrefs found Google's AI Mode and AI Overviews cite different sources 87% of the time for the same query. So "Google AI" is not one surface. Check that the tool separates them if Google matters to you, and that every engine is reported separately. Why engines disagree about your brand covers the causes.
Prompt volume
Count prompts, not features. A prompt set must cover the categories you sell in, the stages of the buying journey and your competitors. Fifty well-chosen prompts beat 500 vague ones. Ask whether the tool counts one prompt across several engines as one prompt or several; that changes what a plan limit buys. Build the set with the free AI Prompt Finder and the guidance in Which prompts should you track?
Citation data
Mentions and citations are different outcomes (see citation vs brand mention). A tool should report both, and show the source URLs so you can see which third-party pages shape the answer. Ahrefs suggests brand mentions may make a better KPI than citations because they are more stable, so a tool that reports only citations is half a measurement.
Attribution
No answer-tracking tool can prove that a named brand caused a sale. Look for the inputs: which AI crawlers fetch your pages, which assistants send visitors, and an API or export to join with your analytics. The three-layer approach is in AI Search Attribution.
Reporting
Your report has two audiences: the person who fixes pages and the person who funds the work. The first needs per-engine, per-prompt detail. The second needs a trend line with a stated method. Check that whatever export and reporting the tool offers works, and that definitions of mention rate and share of voice are written down.
What should you test before buying an AI visibility platform?
Run these six tests during any trial. Each is short, and each produces evidence you can show a colleague.
- The ground-truth test. Pick five prompts. Ask each engine yourself in a clean session, then compare with what the tool stored. If the tool says you were named and your own run disagrees on the same engine, ask why. Variation is normal; systematic disagreement is not.
- The brand-name test. Add alternate spellings and your product names. Check that the tool catches "Acme" in "Acme Inc." and does not match a different company with a similar name.
- The competitor test. Add three real competitors. Confirm the tool lists who is named instead of you on prompts where you are absent.
- The source test. Open five cited URLs. Are they real pages, and do they contain the claim the answer made? Sources you do not recognise are worth a second look.
- The repeat test. Check how many times each prompt runs and whether results are shown as a rate over runs rather than one yes or no.
- The export test. Pull the data out in whatever form the tool offers. If you cannot get answers, mentions and sources out, you cannot join them to analytics later.
For the "known wrong" case, include one prompt where you know the brand should appear, such as your own name. If a tool misses it, it will miss harder ones.
How to use the Noma by Nurdd 3-day free trial as a test
Noma by Nurdd offers a 3-day free trial with 25 tracked prompts. A card is taken at signup so billing can start on day 3; cancel before then and nothing is charged. Three days is short, so plan it. A daily tracker will give you roughly three data points per prompt, enough to test mechanics but not to judge trends. Treat it as an evidence-gathering sprint.
| Day | Do this | What you learn |
|---|---|---|
| Before signup | Run the free brand audit and write down five prompts you checked by hand | Your ground truth, with no tool involved |
| Day 1 | Add your brand, alternate spellings and three competitors. Load 25 prompts: category, comparison, problem and brand-name questions. Read the first answers against your hand checks | Whether brand matching, engine coverage and stored answers hold up |
| Day 2 | Open the cited sources and the gap analysis. List the pages named for prompts you are absent from | Whether citation data points to work you can actually do |
| Day 3, before billing | Check the agent analytics view and the reporting. Decide on a written scorecard, then keep or cancel | Whether reporting and attribution inputs fit your stack |
Keep your prompt set fixed after day 1. Changing prompts mid-trial makes any change in the answers impossible to read; the prompt tracking approach depends on it. Starter is $79 a month for 50 tracked prompts and one project; current plans are on the pricing page.
Where Noma fits
Noma by Nurdd is an AEO and AI visibility platform that asks 13 AI engines, including ChatGPT, Gemini, Perplexity and Claude, the questions your buyers ask every day. For each engine it reports visibility, position and sentiment, compares you with competitors, lists the sources each answer cites, and shows which AI crawlers fetch your pages. Noma launched in May 2026.
See it on your own brand: book a demo or start the 3-day free trial.
What do AEO tools not do?
They do not change your rankings in AI. Measurement only tells you where to work. Google's documentation states there are no additional requirements to appear in AI Overviews or AI Mode, so a tool that promises a special optimisation for them is selling something Google says does not exist. Per Google Search Central, the best practices for SEO remain relevant.
Be equally wary of tools that prescribe one tactic as a guaranteed lever. Search Engine Journal reported an Ahrefs test of 1,885 pages that added JSON-LD schema, matched against control pages, and found no clear citation increase. Schema still helps machines read your pages, but treat every recommendation as a hypothesis and re-test it on your fixed prompt set. See how to get cited more often for work that does move things.
Which tool type fits which team?
| Team | Weight most heavily | First test |
|---|---|---|
| In-house SEO or growth lead | Citation data, source gaps, export | Source test and ground-truth test |
| Agency | Multi-project limits, shareable reports (agency features) | Export test with a real client |
| Founder or small team | Prompt limit, price, setup time | Day 1 setup against five hand checks |
| PR and brand team | Sentiment, description accuracy, cited publications | Competitor and source tests |
Common mistakes when buying
- Buying on engine count alone, without checking that your buyers use those engines.
- Reading one day's visibility score as a trend.
- Skipping the hand check, then trusting a dashboard you never verified.
- Treating AI "position" as a precise rank.
- Expecting the tool to prove revenue, then abandoning it when it cannot.
How we researched this
We reviewed platform documentation from Google Search Central, published research from Ahrefs and SparkToro as reported by Search Engine Journal, and an overview from AirOps, then read Noma's own product facts. All facts, figures and prices were checked on 2026-11-02. We have not tested the tools of other vendors, and this post contains no Noma customer results.
What to do next
Run the free brand audit, write your five hand-checked prompts, then score two or three tools with the checklist above. To test Noma by Nurdd on your own brand, book a demo or start the free trial.
See what AI engines say about your brand
Free · no account · or start a 3-day trial
Questions people ask
What are AEO tools and how do I choose one?
AEO tools ask AI engines your buyers' questions and record whether your brand is named, how it is described and which pages are cited. Choose by engine coverage, prompt volume, citation data, attribution inputs and reporting. Then test the shortlist on your own brand with prompts you have already checked by hand.
What should I test before buying an AI visibility platform?
Compare stored answers with your own manual runs, check that brand names and competitors are matched correctly, open the cited sources, confirm each prompt is repeated, and test whatever export the tool offers. Each test is short and produces evidence you can share.
How many prompts do I need in an AEO tool?
Enough to cover your categories, buying stages and competitors, which for many teams is a few dozen well-chosen prompts rather than hundreds. Noma by Nurdd's trial includes 25 tracked prompts, and the Starter plan includes 50. Check how a tool counts one prompt across several engines.
Can AEO tools prove that AI answers drive revenue?
No tool can prove that a named brand caused a sale. They supply inputs: mention rates, cited sources, crawler activity and referral visits. Join those with analytics and a self-reported attribution field to build the case.
Is a 3-day free trial long enough to evaluate an AEO tool?
It is enough to test the mechanics: brand matching, engine coverage, stored answers and sources. It is not enough to judge trends. Noma by Nurdd's trial takes a card at signup so billing can start on day 3; cancel before then and nothing is charged.
Do AEO tools improve my visibility in AI answers?
They measure it and show where to work. Google's documentation says there are no additional requirements to appear in AI Overviews or AI Mode, so improvement comes from clear pages, third-party coverage and re-testing changes on a fixed prompt set.