What a Drupal GEO & AEO audit covers
Generative Engine Optimization (GEO) is about being cited as a source when an engine composes an answer: the unit of success is a citation, not a ranking position. Answer Engine Optimization (AEO) is about a passage of your content being the answer, self-contained, directly responsive to one question, and usable without the rest of the page around it.
The two lenses fail independently. A page can be citable and still go unquoted because no passage on it stands on its own, and a model-ready answer can sit on a page no engine can reach. Both sit on top of SEO fundamentals, since an engine that cannot crawl, render or index a page will not cite it either, so the audit verifies those fundamentals only to establish whether they are adequate for GEO and AEO purposes.
The audit runs 90 individual checks across access, indexing, rendering, structured data, entity clarity, content shape, readability and AI citation behaviour. Every check carries a topic tag (GEO, AEO, SEO, accessibility or UX) and a scope: the whole website, a set of sample pages, or your target queries.
Coverage is scoped to public-facing pages. Public AI crawlers do not reach authenticated content, so intranets and logged-in areas sit outside the audit by definition. Out of scope by design: technical SEO diagnosis and fixes, which belong to the Drupal SEO audit, plus Core Web Vitals remediation, WCAG conformance targets, copywriting delivery and paid media.
| SEO | GEO | AEO | |
|---|---|---|---|
| What it optimizes for | A ranked link a human clicks | Being cited as a source in a composed answer | Being the passage lifted as the answer |
| Unit of success | Position | Citation | Extracted passage |
| Mostly about | Crawlability, indexing, metadata | Access, authority, entity clarity | Content structure and phrasing |
| How it is measured | Rankings and impressions | Citation rate and share of AI citations | Whether a passage is usable as it stands |
| Who owns it at Metadrop | Drupal SEO audit | This audit | This audit |
A short scoping call separates the SEO problem from the GEO one before you commit to either.
They've Trusted Metadrop
From humanitarian NGOs to industrial manufacturers, organizations across 50+ countries have relied on Metadrop for Drupal audit, structured-data and technical-governance engagements.
Why a healthy Drupal site can still be invisible to AI
Invisibility to AI is a technical failure before it is a content failure. A CDN bot rule, a cookie wall, or content that only exists once JavaScript has run can keep a site out of every AI answer while its Google rankings stay healthy.
robots.txt and the edge configuration can both look correct while every AI bot receives a 403, which is why the audit works from server and CDN access logs rather than from configuration files. There is also no Search Console equivalent that warns you when an engine stops citing you, so the gap can sit unnoticed for months.
Retrieval bots and training bots do different jobs. Retrieval bots (OAI-SearchBot, ChatGPT-User, Claude-SearchBot, PerplexityBot, Bingbot) fetch pages to build a live answer, so blocking them takes the site out of AI answers. Training bots (GPTBot, ClaudeBot, CCBot, Google-Extended) collect content for model training, and blocking them is a legitimate choice with no effect on citation. Google-Extended in particular is not an AI Overviews switch: it governs Gemini grounding and training, while inclusion in Google AI Overviews follows Googlebot plus the snippet directives (nosnippet, max-snippet, data-nosnippet). The audit runs against your stated bot policy per family, recorded in scope before anything is measured, rather than against a house preference.
Engines also state wrong facts about brands with complete confidence: stale prices, closed offices, discontinued services. The source is usually a third-party page nobody has reviewed in years, and tracing where the error came from is often the highest-value output of the engagement.
On the direction of travel, Metadrop cites the Reuters Institute Digital News Report rather than a single headline percentage, because the share of journeys that end inside an AI interface varies by sector and is still moving. Metadrop has also published technical guidance on AI-driven bot traffic and how Drupal-level, CDN and reverse-proxy caching reduce its operational cost, which is relevant here because unmanaged crawler access is a load problem as well as a visibility one.
What Metadrop measures, and what you can re-run yourself
Metadrop reports measured citation behaviour, not a vendor "AI visibility score". Every figure traces back to a query that was run, on a named engine, on a named date, against a named competitor set, so your own team can re-run it and the next cycle stays comparable.
The baseline is anchored to roughly ten target queries you agree up front, spanning your main page types. Changing that set later re-bases the trend, so it is settled in writing. Citations and mentions are counted separately: a citation names or links the source page, while a mention names your brand in the answer body with no attribution. Both are wins, and conflating them inflates the number.
The consultancy phase re-measures a fixed set of quantified trends: AI Overview citation rate, AI assistant citation rate, share of AI citations against your competitor set, entity coverage, AI bot crawl coverage and backlink authority. Alongside them sit the pass/fail signals re-verified each cycle, which are structured-data validity, indexing in Google and Bing, bot access, content originality and freshness, answer readiness, and brand accuracy.
One limit, stated plainly: no agency can promise that a given engine cites a given page for a given prompt. What Metadrop commits to is a measured baseline, a documented method, and a cadence you can read a trend from. Citation behaviour also moves slower than rankings and is noisier, because engines change models and the same query can return different sources a week apart. A cadence of at least a month is the working assumption, and one cycle is not evidence of anything.
Metadrop runs the first measurement, documents the method, and leaves you able to repeat it.
How the audit runs
Scope definition
You and Metadrop agree the AI engines to test, the bot policy per bot family, the roughly ten target queries, the competitor set, and at least one sample URL per page type. All five are conversations rather than defaults. Access is a client responsibility with a date attached: the audit needs Google Search Console, Bing Webmaster Tools and server or CDN access logs, and the logs are what turn "we think we are blocked" into evidence. That request goes into the scope with a deadline.
Audit execution
Crawl-based checks cover every crawlable page, sample URLs are reviewed by hand, bot access is requested per named user-agent and cross-read against the logs, structured data is validated against current schema.org and Rich Results requirements, and the target queries are run against each in-scope engine. The audit reads, it does not change. Findings come from crawls, logs and engine queries against your existing platform, and CMS write access is requested only if remediation is contracted separately.
Analysis and prioritization
Every finding carries a criticality on Metadrop's four-level scale: Very High for a blocker, meaning the goal cannot be met while it stands; High for significant impact, fixed in the same round; Medium for worth fixing, with nothing depending on it; and Informational for a recorded decision or state, reported and not actioned. There is deliberately no "Low" level.
Each finding is also mapped to its Drupal-level root cause and carries an effort estimate, so the recommendation list doubles as the scope of a second phase. Each lens then gets a one-word readiness verdict, Ready, Partial or Not ready, with the findings behind it carrying the detail.
Presentation
Findings are walked through in impact order, in one session with technical and marketing stakeholders together, followed by the prioritized and estimated recommendation list. Audits are timeboxed: the scope is worked to an agreed budget with even depth across sections, rather than one area consuming the engagement.
What you get
Audit deck
Readiness summary, scope and method, baseline metrics, findings by area, prioritized recommendations, and proposed next steps.
Recommendations spreadsheet
Every recommended action ranked by impact against effort, with an estimate, so the list can be scoped as a project or as evolutives inside a maintenance contract.
AI citation baseline
One row per target query per engine, recording who was cited, which page type, and in what position.
AI bot crawl summary
Per bot, read from the logs: requests, unique URLs, the status-code split, and which priority pages were reached.
AI brand accuracy log
Every factual error an engine states about your brand, with the correct fact and the source the error appears to come from.
Consultancy reports
When the ongoing phase is contracted, the same measurements at the agreed cadence, against baseline and target, plus a log of what was implemented.
What makes this a Drupal-native audit
Every finding traces to the specific Drupal module or configuration responsible, whether that is Metatag, Schema.org Blueprints, a robots.txt override, a view mode or the caching layer, instead of a generic symptom description. Where an llms.txt file would help, Metadrop can configure the llms.txt module in Drupal using the platform's native Token system, so the file stays current without manual upkeep. Metadrop has published technical guidance on the llms.txt standard covering its structure, its content-selection criteria and that Token-system implementation.
Findings account for Drupal's multilingual and multisite architecture, including language negotiation, path-prefix setups and domain-based multisite, since AI crawlers encounter each language variant independently. Metadrop also develops and maintains content_first, a Drupal module that is itself one of the audit's tooling groups for content and structure review.
Remediation capability is in-house. Structured data, server-side rendering, edge rules and caching are all work Metadrop delivers, so recommendations arrive with an idea of what they cost to build. Metadrop is a Drupal Silver Certified Partner with 15+ years of production Drupal experience across 50+ countries and 30+ languages, is ENS certified, and carries the same GDPR and WCAG delivery discipline into this audit's own reporting.
Metadrop has deployed and maintained Schema.org structured data on enterprise multilingual Drupal platforms across roughly ten client engagements, covering the Article, Organization, Person, Event, QA Page and Web Page types.
Crawl-control work on high-traffic Drupal sites has included robots.txt governance and a custom query-parameter ignore module, built for duplicate-URL control on an international e-commerce platform.
A Drupal 9 to 11 news-platform migration was built to preserve content, URLs and search equity, with redirect and path-alias migrations handled as first-class migration steps.
Who this audit fits, and who it doesn't
The audit is a good fit when:
- analytics show no referral traffic from AI answer engines, despite reasonable organic rankings;
- a competitor is regularly cited by ChatGPT or Perplexity for queries your organization should be winning;
- an engine states something inaccurate or stale about your organization, and nobody knows where it came from;
- nobody has verified crawler access, so it is unclear whether named AI bots reach and parse your key pages, and nobody has read the logs;
- a relaunch or platform migration is planned, and AI readiness is cheaper designed in than retrofitted;
- leadership is asking for a GEO or AEO position, and marketing needs a technical baseline to answer from;
- a technical SEO or accessibility audit already ran, and the AI layer is the remaining unaudited one.
It is the wrong moment, or the wrong service, when:
- the site fails basic crawling and indexing today, which is worth fixing with the Drupal SEO audit first, because an engine that cannot index a page will not cite it either;
- the platform is an intranet or otherwise sits behind authentication, which public AI crawlers do not reach;
- you want a monthly rank-style score, rather than a method, a baseline and a trend to read over months;
- content writing is the gap you need staffed, since the audit reviews content structure and recommends changes without staffing a copywriting team.
Tell Metadrop your site's scale, language count and current AI-referral signals, and fit gets confirmed before anything is scoped.
Frequently asked questions
Are GEO and AEO the same thing?
GEO and AEO are two lenses on the same platform that fail independently, not two names for one discipline. GEO asks whether an engine can reach, trust and cite you; AEO asks whether a passage of your content works as the answer on its own. Metadrop tags every finding as one or the other, and rates readiness per lens.
How is a GEO & AEO audit different from a technical SEO audit?
A technical SEO audit targets crawling, indexing and ranking signals: sitemaps, canonicals, redirects, metadata, Core Web Vitals. A GEO and AEO audit targets whether AI engines retrieve, quote and correctly describe you. The two overlap on structured data and clean semantic HTML, and SEO fundamentals are the precondition for both. The crawlability and indexation layer is owned by Metadrop's Drupal SEO audit.
Should we allow or block GPTBot, PerplexityBot and Google-Extended?
Whether to allow or block a named AI bot depends on the job that bot does, and the decision is yours per family. Blocking retrieval bots (OAI-SearchBot, ChatGPT-User, Claude-SearchBot, PerplexityBot) takes the site out of AI answers. Blocking training bots (GPTBot, ClaudeBot, CCBot, Google-Extended) is a legitimate choice with no effect on citation. AI crawler traffic has grown across the industry and can add measurable server load, so the audit reviews rate limiting, CDN or reverse-proxy caching and Drupal-level configuration, so that access and performance can coexist. The audit records your intent, then verifies the configuration against it.
What is llms.txt, and does Metadrop implement it in Drupal?
llms.txt is a proposed standard: a markdown file at the site root that gives AI models a structured, low-noise summary of a site's key content. The audit checks whether one exists, whether it is reachable, and whether adding it would help this particular site. Where it would, Metadrop can implement it with the llms.txt Drupal module using the native Token system, so it stays current as content changes. An llms-full.txt variant is evaluated separately rather than assumed.
Why isn't our Drupal site appearing in AI Overviews or chatbot answers?
A Drupal site missing from AI Overviews and chatbot answers usually has one of a handful of causes: blocked or rate-limited AI user-agents, an edge or WAF rule returning 403 to bots, a cookie wall, core content that only renders after JavaScript, incomplete structured data, or no self-contained answer passage on the page. Google AI Overviews inclusion follows Googlebot plus the snippet directives, not Google-Extended. The audit tests access per named bot, cross-reads the server logs, and maps each finding to its Drupal-level root cause.
An AI engine states wrong facts about our organization: what can be done?
When an engine states a wrong fact about your organization, the audit records it in an AI brand accuracy log, with the correct fact and the source the error appears to come from. Errors usually trace to a third-party page, a stale directory listing, or inconsistent naming of the same entity across your own site. Fixes typically combine entity consistency, structured data, public-record presence such as Wikidata, and correcting the third-party source where that is possible. Correction is not instant and depends on the engine re-fetching the page, which is why brand accuracy is a tracked signal rather than a one-off fix.
How long before AI citation behaviour changes?
AI citation behaviour changes more slowly than search rankings, and more noisily: engines change models, and the same query can return different sources a week apart. Plan on a cadence of at least a month, and read the trend rather than a single measurement, because one cycle is not evidence of anything. Quick technical wins, such as unblocking retrieval bots or fixing a rendering dependency, show up in the crawl data sooner than in citation counts.
Request your GEO and AEO audit
Tell Metadrop your Drupal site's scale, its languages, and what prompted the question: a competitor showing up in AI answers, a flat AI-referral number, an inaccurate fact you spotted, or an upcoming relaunch.
Request your GEO and AEO audit
Tell Metadrop your Drupal site's scale, its languages, and what prompted the question: a competitor showing up in AI answers, a flat AI-referral number, an inaccurate fact you spotted, or an upcoming relaunch.