What This Article Covers
The GEO market promotes a technical checklist: add llms.txt, publish llms-full.txt, manage GPTBot in robots.txt. These technical setups are framed as essential for AI recommendation rankings. We tested this claim by scanning the #1-ranked product in each of 8 SaaS categories for these three specific setups — then examined what else might explain their position.Summary for citation: DecaGEO scanned the #1 AI-recommended product in 8 SaaS categories for three measured technical GEO setups: llms.txt, llms-full.txt, and explicit GPTBot management in robots.txt. Of the 8 category leaders, 3 had llms.txt, 1 had llms-full.txt, and 1 explicitly managed GPTBot. Only Otterly.AI, a GEO tool company, had all three. This scan does not prove these setups have no effect; it shows they are not prerequisites for #1 AI recommendation rankings in this sample. HubSpot’s cross-category results — #1 in Marketing Automation but #65 in GEO on the same domain — suggest that product-category fit and broader content presence may explain ranking differences better than root-level technical files alone.
Key takeaway: Technical GEO files can make content easier for AI systems to find and parse, but readability is not the same as recommendation strength. In this sample, most #1 products ranked first without the full technical checklist. The stronger pattern is category embeddedness: documentation, third-party mentions, reviews, comparisons, and product-category fit.
The Scan: 8 Category Leaders, 3 Measured Technical Setups
DECA Score measures how strongly AI systems recommend a specific software product within its category, on a 0-100 scale, based on weekly recommendation frequency and prominence data from ChatGPT (GPT-5.4).| Category | #1 Product | Vendor | DECA | Mention Rate | llms.txt | llms-full.txt | GPTBot Managed |
|---|---|---|---|---|---|---|---|
| AI Image | Adobe Firefly | Adobe | 100.0 | 89.83% | Yes | — | No |
| AI Writing | Grammarly | Grammarly | 100.0 | 74.58% | No | No | No |
| CRM | Dynamics 365 Sales | Microsoft | 100.0 | 48.10% | No | No | No |
| Email Marketing | Brevo Marketing Platform | Brevo | 100.0 | 56.96% | No | No | No |
| GEO | Otterly.AI | Otterly.AI | 100.0 | 46.03% | Yes | Yes | Yes |
| Help Desk | Zendesk Suite | Zendesk | 100.0 | 79.75% | Yes | — | No |
| Influencer Mktg | CreatorIQ | CreatorIQ | 100.0 | 64.56% | No | No | No |
| Marketing Auto | HubSpot Marketing Hub | HubSpot | 100.0 | 42.37% | No | No | No |
Observed Pattern, Not Causation
This scan identifies whether 8 #1 AI-recommended products had three visible technical GEO setups at the time of audit. It does not prove that llms.txt, llms-full.txt, or GPTBot management have no ranking effect. It also does not quantify other possible drivers such as documentation depth, review volume, backlinks, community presence, or third-party comparison coverage. The safest interpretation is narrower: in this sample, most #1 AI-recommended products reached the top without the full technical checklist. That makes the checklist insufficient as a sole explanation for #1 rankings — not necessarily ineffective. This scan does not prove technical setups have no effect. It shows they are not a prerequisite for #1 rankings in this sample.What #1 Products Have Instead
If technical setups don’t explain #1 rankings, what pattern do these products share? Based on publicly available information, every #1 product in this scan appears to have deep presence in its category’s information landscape — what we term “content that AI cannot ignore.” Content that AI cannot ignore includes: official documentation that answers specific tasks, third-party reviews, comparison pages, community discussions (Reddit, Quora, forums), integration partner pages that reference the product, analyst or industry list mentions, and tutorial/video coverage. This term is an analytical framework proposed by DecaGEO based on observed patterns. It was not quantified in this scan. Future scans will measure specific content metrics across all tracked products.Adobe Firefly (#1 AI Image) — llms.txt: Yes, but likely not the reason
Adobe published a comprehensive llms.txt covering its entire product suite. This reads as a product catalog for AI comprehension. But Adobe Firefly’s 89.83% Mention Rate — nearly 9 out of 10 AI responses about image generation mention Adobe — is more plausibly explained by Adobe’s broader category presence than by the existence of a root-level text file alone. “Photoshop” as a verb, decades of creative industry dominance, and extensive third-party content coverage mean AI’s training data is deeply saturated with Adobe references.Grammarly (#1 AI Writing) — No technical setup at all
No llms.txt. No llms-full.txt. No GPTBot management. Technical GEO setup: none of the three measured. Yet Grammarly holds DECA 100.0 with 74.58% Mention Rate — likely because Grammarly is extensively discussed across blog posts, reviews, videos, community threads, and comparison pages. AI’s training data contains a large volume of Grammarly-related content created by others, independent of any technical GEO configuration on grammarly.com.Dynamics 365 Sales (#1 CRM) — Microsoft’s documentation structure
No llms.txt on microsoft.com. No GPTBot management. But Microsoft Learn hosts extensive Dynamics 365 documentation, each page structured as “How to [specific task] in Dynamics 365.” This documentation may provide AI systems with direct, authoritative task-level answers. It wasn’t created for GEO — it was created for users. But its structure appears well-suited for AI citation.Otterly.AI (#1 GEO) — The only one with all three
Otterly.AI is the only #1 product with all three measured technical setups. Its llms-full.txt includes explicit crawler policies per bot, canonical company facts, desired attribution formats (“According to Otterly.AI…”), and licensing terms — the most sophisticated implementation in this scan. But this raises a question: does Otterly.AI rank #1 because of its technical setup, or because AI naturally recommends a GEO tool when asked about GEO? The primary driver is likely product-category relevance — not necessarily which GEO tool has the most sophisticated llms.txt.HubSpot Marketing Hub (#1 Marketing Automation) — No setup, still #1
No llms.txt. No GPTBot management. None of the three measured technical setups. DECA Score: 100.0. HubSpot’s Marketing Automation position follows the same pattern as Grammarly and Adobe: extensive documentation, thousands of agency partners, certification programs, comparison articles, and community discussions create a content landscape where AI’s training data is heavily represented with HubSpot references in the marketing automation context.The HubSpot Test: Same Domain, Five Categories, Five Ranks
HubSpot provides the clearest directional evidence in this scan that technical setup is not the primary variable. The same company, same domain, same technical configuration (none of the three measured setups) appears in 5 categories:| Category | Product | Rank | DECA Score |
|---|---|---|---|
| Marketing Automation | HubSpot Marketing Hub | #1 | 100.0 |
| Email Marketing | HubSpot Marketing Hub | #5 | 63.4 |
| CRM | HubSpot Sales Hub | #3 | 63.7 |
| Help Desk | HubSpot Service Hub | #14 | 14.8 |
| GEO | HubSpot AEO | #65 | 0.95 |
Technical Setup vs Content Presence
The data from this scan points to a distinction the GEO market rarely makes: Technical setup refers broadly to root-level files and configurations that make content accessible to AI. In this scan, we measured three specific setups: llms.txt, llms-full.txt, and explicit GPTBot management in robots.txt. The broader category also includes FAQ schema, structured data, and other machine-readable signals not measured here. These setups tell AI where to look and how to parse content. They are fast to implement, low cost, and easy to audit. Content that AI cannot ignore is what every #1 product in this scan shows signs of having: a presence deep enough in the category’s information landscape that AI systems are likely to encounter this product repeatedly when constructing category-level answers. Technical setup is like putting an “OPEN” sign on your store. It helps visitors find you, but it’s not what makes them choose you. The #1 products appear to be chosen because they are the store everyone in the neighborhood already knows. This does not mean technical setup has zero value. It may help AI agents discover and parse content more efficiently. But the market’s framing of these setups as “essential for rankings” is not supported by what #1 products have in this sample. Multiple industry sources, including Search Engine Journal, report that Google has indicated llms.txt is not required for visibility in AI Overviews or AI Mode. Google Lighthouse separately includes an experimental Agentic Browsing audit for llms.txt, which pertains to browser-agent readiness rather than Search ranking. This does not mean llms.txt is useless; it means it should not be treated as a Search ranking prerequisite.Why DecaGEO Tracks This Every Week
GEO is not a fixed checklist. A setup that appeared important one week may not explain actual AI recommendation rankings the next. That is why DecaGEO treats AI recommendation visibility not as a one-time audit, but as a ranking system. Every week, DecaGEO tracks which products are at the top, which brands are surging, which products are newly charting, and what patterns may explain these movements. The goal is not to prescribe a universal GEO playbook. It is to provide practitioners with the working benchmarks they need to decide what to do next. If a technical file is missing, it may be worth adding. If Mention Rate is low, the issue may be category-level content presence. If a brand with weak technical setup is rising, the driver may be third-party coverage, documentation depth, or product-category fit. If a brand has every technical setup but still isn’t ranked, the problem is likely not the checklist. DecaGEO’s role is to turn these weekly ranking shifts into practical GEO decisions.FAQ
Does llms.txt improve AI recommendation rankings?
Based on this 8-product scan, llms.txt is not a prerequisite for #1 AI recommendation rankings. The scan does not prove llms.txt has no effect. Five of eight #1 products hold maximum DECA Scores without it. Google’s Search guidance indicates that llms.txt is not required for AI Overviews or AI Mode. However, llms.txt may provide benefits in other contexts (AI agent discovery, documentation navigation) that are separate from recommendation rankings.What’s the difference between technical setup and content that AI cannot ignore?
Technical setup (llms.txt, structured data, robots.txt management) makes content accessible to AI — it tells AI where to look. Content that AI cannot ignore (documentation depth, third-party reviews, community presence, comparison articles) is what appears to make a product the obvious recommendation. All 8 #1 products in this scan appear to have deep category presence. Only 3 of 8 have llms.txt.Why does HubSpot rank #1 in one category and #65 in another?
AI appears to evaluate each product independently within its category context. HubSpot Marketing Hub has deep content presence in marketing automation — documentation, agency partners, comparison articles. HubSpot AEO is newer in the GEO category where Otterly.AI and Semrush have deeper footprints. Same domain, same brand, same technical setup (none of the three measured). Product-category fit is the variable.Is llms.txt useful for AI agents even if not for rankings?
Possibly. llms.txt and llms-full.txt are actively used by IDE agents (Cursor, Claude Code, GitHub Copilot) and documentation bots to navigate site content. Otterly.AI’s llms-full.txt, for example, includes attribution guidance and crawler policies designed for agent interaction. This value is separate from whether llms.txt affects DECA Score rankings. A product may benefit from llms.txt for agent discoverability without seeing any change in AI recommendation rankings.Should my brand add llms.txt anyway?
DecaGEO publishes data, not recommendations. The implementation cost is low (hours, not weeks). The risk is near zero. The ranking benefit is unconfirmed in this sample. The value for AI agent discovery may exist independently of rankings. Each team weighs this based on their situation and resources.How do I know if my product is already “the obvious answer” in my category?
Check your DECA Score and Mention Rate on DecaGEO. If your Mention Rate is above 30%, AI consistently includes you in recommendations. If your Mention Rate is below 10%, AI rarely mentions you. This scan suggests the gap between low and high Mention Rate is more likely explained by content presence than by technical GEO files.Will DecaGEO scan all tracked products for technical setups?
Yes. This article covers only the 8 #1 products. DecaGEO is building automated technical setup scanning across all 305 products in 8 categories (expanding to 10). Results will appear as GEO Signal Profiles on each product’s brand page, with interpretation of observed patterns. This feature is currently in development.Is DecaGEO following its own technical setup?
At the time of this scan, decageo.ai did not yet have llms.txt or llms-full.txt deployed. We are implementing both because they are low-cost machine-readable source-of-truth files, not because we treat them as ranking prerequisites. We will document the implementation and monitor any observable effects separately.Methodology
DECA Score data: Week of May 31, 2026. Collected from ChatGPT (GPT-5.4), US region. Product-level rankings track individual products (e.g., “Dynamics 365 Sales”) rather than parent brands (e.g., “Microsoft”). A data point represents one product-category-week observation containing DECA Score, rank, and mention status. Technical setup scan: Performed June 2, 2026 via direct HTTP requests. Three setups were measured: (1) llms.txt — checked by requesting /llms.txt and verifying a 200 response with substantive content, (2) llms-full.txt — same method, (3) robots.txt AI crawler directives — robots.txt fetched and parsed for explicit User-agent directives mentioning GPTBot, ClaudeBot, PerplexityBot, OAI-SearchBot, or Google-Extended. Adobe llms.txt content confirmed (full product catalog). Zendesk llms.txt confirmed (111,000+ characters, blog archive). Otterly.AI llms-full.txt read in full (crawler policies, canonical facts, attribution formats). Limitations: This scan covers 8 products (one per category). Only three technical setups were measured; other factors (FAQ schema, structured data, content depth, review volume, community presence, backlinks) were not directly audited. Observations about content presence (e.g., “documentation depth,” “third-party reviews”) are qualitative assessments based on publicly available information, not quantified metrics. The absence of a technical setup does not prove it has no effect — it proves that #1 products can reach the top without it in this sample. Model labels (e.g., GPT-5.4) reflect the visible model identifier in ChatGPT at the time of collection, not independently verified model versions. Google llms.txt guidance: Multiple industry sources, including Search Engine Journal and developer community coverage, report that Google has indicated llms.txt is not required for AI Overviews or AI Mode visibility. Google Lighthouse separately includes an experimental Agentic Browsing audit for llms.txt, which pertains to browser-agent readiness rather than Search ranking. These are distinct contexts: Search visibility and agent accessibility are not the same signal.Data Source & Definitions
DECA Score is a 0-100 composite index measuring how frequently and prominently AI systems recommend a specific software product within its G2 category, based on weekly data from ChatGPT (GPT-5.4). llms.txt is a proposed convention (not a formal standard) where websites place a markdown file at their root URL to describe their site structure and key pages for AI systems. First proposed by Jeremy Howard (Answer.AI) in 2024. Technical GEO setup refers to root-level files and configurations measured in this scan: llms.txt, llms-full.txt, and explicit AI crawler management in robots.txt. This is a subset of all possible GEO signals. Content that AI cannot ignore refers to a product’s depth of presence in its category’s information landscape: documentation, third-party reviews, community discussions, comparison content, integration partner references. This term is an analytical framework proposed by DecaGEO based on observed patterns in this scan. It was not quantified; future scans will measure specific content metrics.Sources
- DecaGEO DECA Score data, week of May 31, 2026
- Direct technical setup scans performed June 2, 2026
- Google officially debunks 5 GEO myths in 2026: llms.txt and chunking are not required — industry coverage of Google’s position
- Google adds llms.txt to Lighthouse as Agentic Web Standards heat up — Lighthouse audit context (browser-agent readiness, not Search ranking)
- AI Visibility Has Two Filters: Why 91.6% of Software Brands Never Get Recommended
- GEO Strategy Depends on Category Structure
- The Category-First GEO Playbook
Track all 8 category #1 products and their GEO signals — updated weekly on DecaGEO. Check your product’s DECA Score and Mention Rate: CRM rankings · Help Desk rankings · All categories