AEO and GEO framework

aeo-geo-strategy.md: addition-labs.com

Scope of this audit: the homepage (/), /robots.txt and /llms.txt, fetched 2026-09-11 with curl. Every URL listed in llms.txt was HEAD-checked for status. One Wikidata search and one Wikipedia lookup were run for the entity layer. Nothing else on the site was read, so scores describe what these three surfaces show, not the whole site.

1. Current AI visibility audit

Gap, stated per the data-availability rule. The workflow's step 3 requires querying the major AI products (ChatGPT search, Google AI Overviews, Perplexity, Gemini, Claude) with each priority query and recording which sources they cite. That cannot be done from this environment. So:

• Needed: live citations for ecommerce seo agency, aeo agency, what is answer engine optimization from each AI product.
• Obtained: none. No AI product was queried.
• Affected: this section, the "is the site cited" column of the priority-query table in section 3, and the competitor gap analysis (step 4). Those are left empty rather than estimated. Everything else in this document rests on the fetched files and is unaffected.

What can be said without the live test: the site has a dedicated page for each of the three queries (see section 3), all three return HTTP 200, and all three are listed in llms.txt. Whether any AI product cites them is unknown.

2. Five-layer scorecard

Scores are 0 to 10 and are judged on the fetched surfaces only.

Weighted read: the site is well ahead of most agency sites on layer 4 and behind on layers 2 and 5, which are the layers that decide whether an AI product trusts the source enough to name it.

3. Priority queries

The three queries given, mapped to the page that should carry the citation. "Cited today" is unknown for all three (section 1).

The skill asks for 10 to 20. The following are proposed from pages that already exist in llms.txt; they are candidates for the owner to confirm, not measured demand:

1. what is ai visibility → /resources/glossary/what-is-ai-visibility
2. aeo vs seo → /resources/compare/aeo-vs-seo
3. chatgpt ads agency → /services/chatgpt-ads
4. ecommerce google ads agency → /services/google-ads
5. shopify aeo → /shopify/aeo
6. ai visibility monitoring services → /services/aeo/ai-visibility
7. llm optimization → /services/aeo/llm-optimization
8. b2b aeo → /services/aeo/b2b
9. best ai visibility tools → /resources/tools/ai-visibility-tools
10. answer engine optimization tools → /resources/tools/answer-engine-optimization-tools
11. geo tools → /resources/tools/generative-engine-optimization-tools
12. llms.txt generator → /tools/llms-txt-generator
13. chatgpt optimization → /services/aeo/chatgpt

4. Layer-by-layer remediation plan

Layer 1: extractable structure (homepage, then the three target pages)

1. Add a definitive two-sentence answer under the H1 that says what Addition is in extractable terms (agency type, channels, who it serves). The current subhead ("Your next customer is already looking for what you sell...") is a hook, not an answer.
2. Recast at least two H2s as the questions buyers ask ("What does an AI-native search agency do?", "Which channels does Addition run?") and open each with the answer.
3. Add one comparison table (the four services × what each covers × how it is measured); the "Services. Four ways people find you" block is already that data in prose.
4. Remove carousel affordances from the crawlable text: render the ← → controls and "Swipe for all six steps" as ARIA-labelled buttons, and keep the You/We digit graphic out of the DOM text (use aria-hidden or an image with alt).
5. Not fetched, so not scored: confirm the three target pages open with a direct answer and, for the glossary entry, that the definition is the first sentence.

Layer 2: citation worthiness

1. Put at least one specific, sourced number on the homepage from the case studies the site says it has, with the client named or the anonymisation stated.
2. Add a named author with a bio link on every guide, glossary and comparison page, marked up as Person with sameAs (LinkedIn at minimum).
3. Show publication date and last-updated date on every resource page, visibly and in schema.
4. Where a page states "what it could not measure", keep that; it is exactly the methodology disclosure this layer rewards. Mirror a one-line version of it on the homepage.

Layer 3: structured data

1. Extend Organization: description, telephone (if one exists), foundingDate, founder (as Person), areaServed, knowsAbout (SEO, AEO, Google Ads, ChatGPT Ads), contactPoint, and more sameAs entries (any owned profiles beyond LinkedIn).
2. Add Service entities (one per channel) referencing the organization as provider.
3. Add BreadcrumbList site-wide; Article with author, datePublished, dateModified on resources; DefinedTerm or Article on glossary entries; FAQPage only where a page has genuine Q&A pairs (do not manufacture them).
4. Align og:url with the canonical (https://addition-labs.com/).
5. Validate in both the Schema.org Validator and Rich Results Test after each change.

Layer 4: AI-readable accessibility

1. Cut llms.txt to a curated set: keep the five service pillars, the case studies, the glossary and comparison pages that map to the priority queries, the tools, and the company pages; target 30 to 60 links and under 2,000 words. Move the full inventory to sitemap.xml (already present) or to llms-full.txt if a full dump is wanted. Keep the description; it already names what, for whom and how measured.
2. Make the crawler decision explicit in robots.txt: list GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, Google-Extended, PerplexityBot with Allow: / so the allow is a stated policy, not the absence of one.
3. Add a <nav> element around the primary navigation.
4. Optional: add <link rel="alternate" type="text/markdown" href="/llms.txt"> to the homepage head so the file is discoverable from the HTML.

Layer 5: entity signals

1. Create a Wikidata item for Addition Labs LLC (legal name, website, LinkedIn, address, founding date, industry) with properties matching the JSON-LD.
2. Add every owned profile to sameAs and make the NAP identical across Google Business Profile, LinkedIn, Crunchbase or similar, and the site footer.
3. If clients can be named, have them named consistently on the site, on LinkedIn and in any press, so the entity claims cross-reference.
4. Reviews: choose one platform (Clutch or Google Business Profile are the usual ones for agencies), collect real reviews there, and reference it from the site. Do not add AggregateRating schema without a real, visible rating source.
5. Do not pursue a Wikipedia article unless independent coverage exists.

5. Implementation roadmap

6. Re-test schedule

• Re-run the 16-query visibility test across the AI products at the end of week 6, then quarterly (December 2026, March 2027, June 2027).
• Re-check llms.txt links for 200 status and re-review its contents quarterly, or when a new section launches.
• Re-run this five-layer scorecard quarterly; AI products change faster than search algorithms did, so a score from one quarter is not evidence for the next.

Data gaps, in one place

• No AI product was queried: section 1, the "cited today" column, and the competitor gap analysis are empty.
• Only the homepage, robots.txt and llms.txt were read; layers 1 to 3 are scored on the homepage alone, and the three target pages were checked only for HTTP status.
• Schema was parsed but not run through the Schema.org Validator or Rich Results Test.
• Third-party brand mentions, reviews and social profiles beyond the single sameAs link were not measured.
• Queries 4 to 16 are proposals derived from existing pages, not measured search or prompt demand.