AEO and GEO technical audit
for consumer brands.
An AI assistant can only recommend your brand if it can crawl your site, parse your product data, and trust what it finds. The AEO technical audit checks whether it can. Roughly 50 checks across five areas, delivered as a fix list ordered by impact and effort, for consumer brands in the United States.
This is the plumbing half of AI visibility. The other half is measuring what the assistants actually say about you, which is the AI Visibility Audit. Both are delivered together in one fixed-price engagement, because a technical finding without a prompt result is a guess about what matters.
What we check
Five areas. Every check is pass, fail, or partial, and every failure maps to a specific fix.
1. Bot governance
Whether the assistants are allowed in at all. This is where the most expensive failures hide, because a single line in a config file can make a brand invisible everywhere at once.
robots.txtaccess for GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, Claude-User, PerplexityBot, Perplexity-User, and Google-Extended- Accidental blocking through a CDN, WAF, or bot-management rule that never appears in robots.txt
- An
llms.txtmanifest, present and pointing at clean, current content rather than a stale page list - Sitemap validity, discoverability from robots.txt, and honest
lastmoddates - Server responses to AI crawler user agents, including rate limiting and challenge pages that silently reject them
2. Structured data
Machines prefer facts they do not have to infer. Schema is how you hand them over without ambiguity.
- JSON-LD
Organizationwith completesameAsentity links to your real social and retailer profiles Productmarkup on product pages with GTIN, brand, ingredients, size, and category attributes actually populated rather than stubbedFAQPagemarkup that matches visible on-page content, since schema describing content that is not on the page is a violation and can invalidate the blockOfferdata: price, currency, availability, and whether it reflects reality today- Validity, duplication across templates, and conflicts between competing blocks on the same page
3. Content structure
Assistants quote sentences, not pages. The question is whether your page contains a sentence worth quoting.
- Definitional first sentences: does the page answer its own question in the opening line, or bury it under brand voice
- Semantic heading hierarchy that describes content rather than styling it
- Answer-ready FAQ coverage of the need-state questions buyers actually ask, matched against your real prompt results
- Product facts stated in extractable text, not locked in images, PDFs, or JavaScript-rendered components
- Claim substantiation near the claim, so a model has something to cite when it repeats you
4. Entity strength
Models trust brands they can identify consistently across the web. Inconsistency reads as uncertainty.
- Consistent brand naming across your site, retailer product pages, social profiles, and press
- Wikidata and Wikipedia presence where the brand qualifies, and accuracy where it already exists
- Third-party citations and earned mentions in the sources models draw on for your category
- Founder and company entity linkage, so the people and the brand resolve to each other
- Conflicting facts across the web that give a model reason to hedge or omit you
5. Commerce data readiness
The check most SEO audits skip entirely, and the one that decides whether a shopping agent can transact with you.
- PDP attribute depth against what agents filter on: dietary flags, ingredients, certifications, pack size, use occasion
- Real-time pricing and availability, and whether your feed agrees with your page
- Return and shipping policy clarity, which agents score merchants on directly
- Retailer listing hygiene on the sites models cite most in your category
- Review platform presence and whether the review content is machine-readable
What you get back
The fix roadmap
Every failed check mapped to impact and effort, with the top ten sequenced and quick wins flagged. Written conclusion-first, in the form: you are invisible for these category queries because of this, fix that first. Your team can execute it without us, and plenty do.
The one-page scorecard
A single page built to be forwarded internally without translation, because it will end up in front of your CEO. Five area scores, the headline finding, and the three things that matter most.
The evidence
Every check result, with the specific URL, the specific line, and what we found. No claim in the report exists that you cannot verify yourself in five minutes.
Implementation, if you want it
Agentic Commerce Readiness quotes directly from your own roadmap, $3,000 to $8,000 per scope, so there is no second discovery phase to pay for. Optional, and a fair number of brands skip it.
See GEO consulting →Scope note: the technical audit is delivered inside the $1,950 AI Visibility Audit rather than sold on its own. That is deliberate. A technical checklist in isolation produces a list of fifty things that are all technically true and gives you no way to know which three matter for your category. Running the prompts first is what makes the list rankable.
Common questions about AEO technical audits
What is an AEO technical audit?
An AEO technical audit, for answer engine optimization, checks whether a site is machine-readable enough for AI assistants to crawl, parse, trust, and quote. It covers five areas: bot governance, structured data, content structure, entity strength, and commerce data readiness. Shelf Help AI runs roughly 50 checks and returns a prioritized fix roadmap rather than a raw export. It is included in every AI Visibility Audit at $1,950 fixed.
How is an AEO technical audit different from a technical SEO audit?
They overlap on fundamentals: crawlability, site speed, clean markup, valid sitemaps. They diverge on what they optimize for. A technical SEO audit is trying to get a page ranked and clicked. An AEO technical audit is trying to get a fact extracted and quoted, so it weighs structured data completeness, extractable product attributes, definitional writing, and entity consistency far more heavily. It also includes checks a traditional SEO audit skips entirely, such as whether GPTBot and ClaudeBot are actually admitted and whether your commerce data is legible to a shopping agent.
Can I just run a free tool instead of paying for an AEO audit?
Partly, and you should. A schema validator will tell you whether your JSON-LD parses. Reading your own robots.txt will tell you whether GPTBot is blocked. Those two checks are free and catch the most expensive failures, so do them today whether or not you ever hire anyone. What a validator will not do is tell you which of fifty valid-but-incomplete findings actually explains why a competitor is winning your category queries. That prioritization is the part that needs the prompt results next to it.
What is llms.txt and do I need one?
An llms.txt file is a plain-text manifest at the root of a site that tells AI systems what the site is, what it offers, and which pages carry the substantive content. It is a proposed convention rather than a standard that every assistant honors, so treat it as low-cost insurance rather than a fix. It takes an hour to write, costs nothing to host, and helps the assistants that do read it. Shelf Help AI publishes one at shelfhelpai.com/llms.txt if you want to see the format.
How long does an AEO technical audit take?
About one week for the full engagement, which includes the prompt testing and the technical audit together. No call is required to start, though a 30-minute readout is available when the report is delivered.
Will you implement the fixes or just list them?
Either. The audit ends with a roadmap your team can execute, and many brands do exactly that, which is the intended outcome. If you would rather have the work done, Agentic Commerce Readiness handles implementation at $3,000 to $8,000 per scope, quoted from the roadmap the audit already produced.
Check the cheap things first.
Before buying anything, open your own robots.txt and search for GPTBot. If it is disallowed, you have found your problem and you did not need us. If it looks fine, the free AI Shelf Snapshot will tell you whether the assistants are actually recommending you.
Get Your Free SnapshotReady for the full diagnosis? See the AI Visibility Audit →