Deep dives
Best AEO tools for agencies (2026)
The stack an agency runs to deliver answer engine optimization for clients: what each tool does, what it costs, and where the free ones are already enough.
An agency selling AEO does not need a shelf of tools that do the same thing. It needs one tool that tracks and audits a client domain and reports in the agency's name, plus a few specialists that answer the questions the audit raises.
Full disclosure before the list: Maverank is our product, and it covers three of the eight jobs below. Every price and free tier was checked on the vendor's own pages on 2026-07-25.
What an agency AEO stack has to cover
Eight jobs, and the client only ever sees the last one.
- Baseline. Read what the search engines already record about the site, from inside their own consoles rather than from outside.
- Submit. Get new and changed URLs into the indexes fast, rather than waiting for a crawler to notice.
- Validate. Check the machine-readable parts, structured data and llms.txt, against the spec rather than against a plugin's promise. Keep the llms.txt half in proportion, because no major provider documents reading the file.
- Observe. See which AI crawlers actually reached the site, as opposed to which ones the robots.txt says are welcome.
- Track and audit. Ask the engines the client's buyer questions on a schedule, and audit the site for the reasons behind the answers.
- Fix. Implement the changes in the client's repository, rather than emailing a list and hoping.
- Prove. Show that the fixes moved something, on both the answer side and the traffic side.
- Report. Deliver it under the agency's logo, on a schedule, without paying per client for the privilege of showing them their own data.
Six of the nine tools below are free. The baseline consoles are the ones most often left half-configured on a client account, and job 6 is the one that quietly eats the retainer.
Both search consoles as the free baseline
Google Search Console and Bing Webmaster Tools are free, and they are the only tools in this stack that report from inside a search engine rather than from outside it. Set both up on every client property on day one, because the historical data only starts accumulating once you do.
Google Search Console tells you what Google itself records about the site. For AEO work it earns its place in three ways:
- index coverage tells you which pages Google can actually see
- sitemap submission tells you what you have declared
- the rich results reporting tells you which structured data Google recognised rather than which you intended
It is not an AI answer tool and does not claim to be.
Bing Webmaster Tools is the same idea on the other index. Microsoft's own list of what it covers includes:
- search performance reporting
- a URL inspection tool
- IndexNow integration and URL submission
- a site scan
- a robots.txt tester
- crawl control
Since February 2026 it also carries an AI Performance report, in public preview, which shows when a site is cited in AI-generated answers, which URLs were referenced, and how citation activity moves over time.
Read the scope before you put it in a client report: Microsoft describes the coverage as Microsoft Copilot, AI-generated summaries in Bing, and unnamed partner integrations. It is not a view of ChatGPT, Claude or Perplexity, and Microsoft does not claim it is.
IndexNow for getting changes seen quickly
IndexNow is a free, open-source protocol in Microsoft's own words, for telling search engines that a URL has been added, updated or deleted instead of waiting to be crawled. Implementation is a hexadecimal key in a text file at the site root, then either a one-URL ping or a POST of up to 10,000 URLs. Submit once to any participating endpoint and the submission is shared with the others, so there is no per-engine integration to maintain.
The live registry lists seven participants: Bing, Yandex, Seznam, Naver, Yep, the Internet Archive and Amazon's crawler. Google is not among them, and Cloudflare is the one CDN the protocol's own FAQ names as offering a native integration.
The honest limit matters more here than the setup, because this is where the category oversells. No AI answer engine participates in IndexNow.
The one defensible AI benefit is indirect and runs through Bing: Copilot citations are what Bing Webmaster Tools reports on, so faster Bing indexing is the route to faster Copilot pickup. Nothing published by OpenAI, Anthropic or Perplexity says they consume IndexNow submissions, and any tool selling IndexNow as an AI visibility lever is ahead of the evidence.
Free validators for structured data
Two free tools cover structured data, and an agency should run both because they answer different questions.
Google's Rich Results Test takes a URL or a code snippet and reports which rich results Google can generate from the structured data on the page. That is Google's requirements, not the vocabulary's.
The Schema.org Markup Validator, run by Schema.org itself, checks markup against the Schema.org vocabulary with no sign-in and no fee.
Valid markup that Google will not use is a real state, and so is markup Google accepts that says something the client did not mean. Running both is how you tell those two apart in a client report.
Cloudflare AI Crawl Control for what actually reached the site
Cloudflare's AI Crawl Control shows which AI crawlers requested a site and lets you allow or block them individually. It is available on all plans, and on the free plan crawlers are identified by user agent string with the metrics view limited to 24-hour windows. Blocking an individual crawler is one action in the crawlers tab, and robots.txt compliance monitoring sits alongside it.
This is the job nothing else in the stack does. A robots.txt that welcomes GPTBot and a CDN rule that returns 403 to it are a common combination, and only the request log settles which one is winning.
Check which bot you are looking at before you report on it: OpenAI's crawler documentation names OAI-SearchBot as the bot used to surface websites in ChatGPT's search features and GPTBot as the one that feeds model training, and states that a site opted out of OAI-SearchBot will not be shown in ChatGPT search answers. Blocking training and losing search visibility by accident is the mistake worth catching on a client account. If the client is already on Cloudflare, this costs nothing and takes one visit to the dashboard.
Maverank for the tracking and the audit
Maverank covers three of the eight jobs: tracking and auditing in the middle, the answer half of the proof at the end, and the report the client actually sees. The first two need each other, which is why they are one section.
The tracking half asks the client's buyer questions across ChatGPT, Claude, Gemini and Google AI Overviews on a weekly or daily cadence, and stores every completed run as a dated snapshot: visibility score, mention rate, average position, sentiment, and share of voice against the competitors you named. That is what makes a trend a trend rather than a screenshot, and it is the same discipline the whole category depends on, because a single run of a non-deterministic engine is an anecdote.
The audit half explains the trend. Every plan runs the same 30 agent readiness checks in four weighted pillars, returned as a prioritized list of what to change, with crawler access and server-rendered content treated as gating checks that cap the score when they fail. Three of the 30 gate the score, because a site that returns 403 to GPTBot or serves an empty shell without JavaScript has one problem and it is not the wording of its headings.
Read-only client viewers are free and unlimited on every plan, including the $59 Solo plan, so adding a client to their own dashboard does not add a seat to the bill. White-label reports and scheduled reports are on the $149 Studio plan, which also carries 10 projects, 200 tracked prompts and API access.
To try it on a client domain before any of that, the free check takes a domain and no credit card and returns the fix list. If you are weighing visibility platforms against each other first, our comparison of five of them covers entry prices, how each one meters prompts, and which engines are on which plan.
Claude Code for implementing what the audit finds
Most of a readiness fix list is not strategy, it is files. A robots.txt rule, an llms.txt file, a block of JSON-LD, a heading rewritten to match the question it answers: small, well-specified changes to a website.
Claude Code is Anthropic's AI coding agent. You hand it the fix list from a report, and it makes the changes in the site's repository and opens them for review.
The division of labour is the point. Tools like Maverank tell you what to fix and why it matters, and a coding agent does the fixing. Knowing what to fix is the hard part nowadays, and that is the half the report covers.
Two working rules if you point an agent at a client site. Have it open a pull request rather than change the live site directly. And re-run the audit after the merge, because the check that failed is the only thing that can tell you it now passes.
Proof: the answer side and the traffic side
The client's question at the end of an engagement is whether any of it worked, and one number cannot answer it.
On the answer side, the record is Maverank's dated snapshots: visibility score, mention rate, average position, sentiment and share of voice against the named competitors, per engine, on every scheduled run. Before and after on a fixed prompt set is a claim an agency can defend, because the prompts, the engines and the competitor set stayed still while the site changed.
On the traffic side, Google Analytics 4 is free, and since 13 May 2026 it has had a built-in AI Assistants channel. GA4 groups visits referred by AI assistants into a built-in AI Assistants channel, so the traffic side of the story is a standard report rather than a custom build.
Read the boundaries before you build a client report on it:
- Google states that the AI Assistants channel excludes its own AI Overviews and AI Mode, which makes the channel a floor rather than a measure of AI-influenced traffic.
- The list of domains Google matches against is not published anywhere we could find.
- Google's own documentation on direct traffic lists the ordinary ways referral data goes missing: redirects, URL shorteners, ad blockers, and links that carry no campaign tags.
Put the two together and the honest sentence to a client is that the engines named them more often on a fixed prompt set, and the assistant referrals GA4 could attribute went up. Neither half is the whole truth alone, and the pair is more defensible than either.
What the stack costs to run
| Tool | Job | Free tier | Paid |
|---|---|---|---|
| Maverank (ours) | audit, fix list, white-label client reports | free check, no card | $59/mo Solo, $149/mo Studio for white-label |
| Google Search Console | Google's own view of the site | free, all of it | none |
| Rich Results Test | structured data as Google reads it | free | none |
| Schema.org Markup Validator | structured data against the vocabulary | free | none |
| Cloudflare AI Crawl Control | which AI crawlers arrived, and blocking | on all plans, 24-hour metrics on free | included with paid Cloudflare plans |
| Bing Webmaster Tools | Bing's view, plus Copilot citations in preview | free, all of it | none |
| IndexNow | instant URL submission to participating engines | free, open protocol | none |
| Claude Code | implementing the fix list in the repository | none we could verify | Claude subscription, or per-token on Anthropic Console |
| Google Analytics 4 | AI assistant referrals on the traffic side | free, all of it | Analytics 360, sold through Google sales |
An agency can run six of the nine at no cost, seven if the client is already on Cloudflare, and still deliver most of the technical half of an AEO engagement. What the paid rows buy is the tracking and the audit, the agent that does the implementing, and a report the client sees with your name on it.
The arc, in order:
- Both consoles for the baseline.
- IndexNow so changes get seen.
- The validators for the standards.
- Cloudflare for who actually arrived.
- Maverank for the tracking and the audit.
- Claude Code for the implementing.
- Maverank's trend and GA4 for the proof.
- A white-label report at the end of it.
Frequently asked questions
What is the minimum AEO stack for a small agency?
Both search consoles, Google Analytics 4, the two free structured data validators and Cloudflare AI Crawl Control cost nothing and cover both indexes, the markup, the crawler question and assistant referrals. Add an audit and reporting tool when you need to hand the client something with your logo on it rather than a spreadsheet, and a coding agent when the fix list stops being something you can action by hand.
Do agencies need a separate tool for white-label client reports?
Not usually, if the audit tool already does it. Maverank puts white-label and scheduled reports on the $149 Studio plan and keeps read-only client viewers free on every plan, so client access and client-facing reports come from the same place the audit does.
Does IndexNow help with AI search?
Only indirectly, and only through Bing. No AI answer engine participates in the protocol, so nothing you submit reaches ChatGPT, Claude or Perplexity by that route. What it does is get a URL into Bing's index faster, and Bing is the index Microsoft Copilot's citations are reported against.
Can you measure traffic from AI assistants?
Partly, and what you get is a floor rather than a total. Google Analytics 4 has had a built-in AI Assistants channel since May 2026, which tags visits whose referrer matches Google's list of assistants. It excludes Google's own AI Overviews and AI Mode, the matched domain list is not published, and the ordinary causes of lost referral data apply here as everywhere else.
Is AEO different from technical SEO?
It overlaps more than the category's marketing suggests. Crawler access, server-rendered content, clean structured data and resolving links are technical SEO work that AI engines happen to depend on, and the genuinely new parts are the machine-readable conventions and measuring whether an engine names the brand at all.
Where to start
Run a free check on one client domain, read the fix list, and see how much of it your current stack was already telling you. If the answer is that client access and branded reporting are the gap, the plans and what each limit means are published in full.