AI Search Readiness Benchmark 2026: A Historical Review of 61 SaaS Sites

The benchmark uses the source repository linked below. That repository is public (access checked October 29, 2026). Read the published methodology and limitations here; local setup time and execution on your site have not been verified.

Want it done for you instead? Book the paid audit (scope and fees agreed after review).

AI Search Readiness is the name of this report’s 100-point inspection rubric. It records selected technical and content checks. The score has not been calibrated against discovery, understanding, trust, citations or rankings, and a higher score does not establish a greater likelihood of any of those outcomes.

AI Search Readiness is the label used here for a historical inspection rubric covering technical access, content and entity checks. The captured results describe the June 23, 2026 v1.12 rescore only. They do not measure an engine’s decisions, establish citation likelihood or demonstrate an optimization outcome.

The legacy rubric groups its checks into the six categories and reported weights below. They are historical checklist categories, not a published weighting model for Google or any answer engine. The captures do not reconstruct the complete June v1.12 scoring algorithm.

1. 01 20 pts
   Technical access

   Have relevant search crawler rules and hosting responses been checked separately from training controls?

2. 02 15 pts
   On-page clarity

   Is the answer to a likely prompt in a self-contained block?

3. 03 20 pts
   Schema.org markup

   Organization, Person, Article, FAQPage, HowTo, Dataset.

4. 04 25 pts
   GEO signals

   /llms.txt, /llms-full.txt, factual statements and original statistics as historical rubric inputs, not citation evidence.

5. 05 10 pts
   AEO signals

   Legacy format checks: short definitions, bulleted lists and numeric evidence; no engine response is established.

6. 06 10 pts
   E-E-A-T

   Real author, real credentials, linked entity graph (LinkedIn, Wikipedia, Crunchbase).

The captured June 23, 2026 v1.12 cohort contains 61 SaaS and AI sites. Its mean rubric score is 54.3/100, median 59/100, and 34 of 61 sites scored 60 or below (56% rounded). OpenAI and Perplexity each scored 7/100. These figures do not identify the causes of current search visibility or citation behavior.

• →56% of top SaaS sites scored 60 or lower. The threshold groups historical rubric scores; it does not classify citation success.
• →Mean: 54.3 / 100. Median: 59 / 100. These summarize the captured rubric, not content quality or search performance.
• →OpenAI and Perplexity both score 7 / 100. A low internal score does not establish invisibility to search or inability to be cited.
• → Per-signal access and schema behavior need separate evidence. A missing optional /llms.txt file does not demonstrate a search blackout.
• →Schema component scores are aggregate rubric outputs. They do not establish whether each page has Organization or Person markup; inspect the relevant page response separately.
• →The captured notes show: /llms.txt appears for 17 of the displayed top 20 sites and 4 of the bottom 10. Other crawler and entity-markup percentages cannot be reconstructed from the captured aggregate scores.
• ◎Historical example: a 94/100 score was claimed for yanivgoldenberg.com on June 5, 2026. That run is unverified here. The captured June 23 v1.12 rescore is 92/100; these are separate records and neither proves current citation performance.

Read the benchmark methodology and the self-audit below to identify questions to investigate. The linked source repository is public; a current score requires a dated run against the intended site.

Want to discuss the questions relevant to your site? Book the paid audit (scope and fees agreed after review).

The captured June 23 v1.12.0 rescore gives yanivgoldenberg.com 92 and Heroku 86. The twentieth displayed row is 65; ties mean the displayed top 20 is not an exclusive top-third threshold. These scores measure the historical rubric, not visibility or citations.

The following table contains the displayed bottom ten sites in the June 23, 2026 v1.12 rescore. OpenAI and Perplexity each recorded 7/100, the lowest composite score in this cohort. These internal rubric results do not establish current crawler access, schema details, content quality or citation outcomes.

The June capture records /llms.txt on 17 of the displayed top 20 sites (85%) and 4 of the bottom 10 (40%). This is descriptive, not causal. The capture lacks the per-signal detail needed to reconstruct the other adoption figures below.

Interpretation. The rubric describes selected technical and content checks. These historical scores do not isolate the causes of search visibility, compare writing quality, establish a required answer length, or measure implementation time.

Use the checklist below to review the technical and content gaps relevant to your site. Treat the benchmark rubric as an inspection aid; the linked script repository is public; its anonymous execution workflow has not been verified here.

Option A: use an authorized repository checkout. Repository access and setup are required. Follow the approved repository instructions for the version you are using; a successful run and its duration still need to be observed.

Review the historical SITES list in tests/benchmark_sites.py to define an intended cohort only after reviewing the current script and method. Record the inputs, run date, version and output; a different cohort is a different measurement.

Option B: review the legacy checklist. Use it to identify questions, not as a measured priority or points-per-minute order.

OpenAI.com scored 7/100 in the captured June 23 rescore. The recorded aggregate components do not establish the homepage’s current crawler access, JavaScript behavior or exact schema types. Absence of optional /llms.txt does not prove that a page cannot appear in search or be cited. The following items are evidence requirements, not verified missing features.

A historical rubric score is distinct from brand strength, product quality, traffic and search outcomes. This cohort does not establish how those variables relate or whether a brand is visible to an answer engine.

The following ideas are hypotheses for investigation, not observed fixes, generated output or predictions. The captured score is historical; the effort, applicability, rubric changes and citation effects have not been measured.

The linked skill describes an audit and possible follow-on work. Its current behavior, generated artifacts, compatibility and claimed before-and-after outcomes have not been verified here. Review access, license terms and the evidence on the skill page.

Robots rules can affect particular crawlers. Search, user-triggered retrieval and training have different controls; one rule does not establish whether any service will cite a page.

The published benchmark reports 61 sites re-scored on June 23, 2026 (v1.12) with a read-only Python script on a 100-point rubric (Technical 20 + On-Page 15 + Schema 20 + GEO 25 + AEO 10 + E-E-A-T 10). Scores are historical outputs of that rubric. The linked repository is public (access checked October 29, 2026); independent reproduction and the full captured run remain separate evidence requirements.

Full raw data

AI Search Readiness

The historical 100-point inspection rubric used in this report. Its technical, content, schema and author/entity categories are checklist inputs, not calibrated probabilities or evidence that an engine can or will cite a page.

GEO (Generative Engine Optimization)

A term used for attempts to improve how content appears in generative search responses. This report does not validate an optimization method or establish a measured result from it.

AEO (Answer Engine Optimization)

A term for organizing content around questions and answers. Clear, supported answers may be useful to readers; no fixed format, numeric detail or word count establishes quotation or ranking.

/llms.txt

An optional convention for exporting site descriptions and URLs for a particular consumer. Google documents no new AI text file requirement for AI Overviews or AI Mode. This benchmark does not establish citation or ranking gains from the convention.

GPTBot

OpenAI’s training crawler. OAI-SearchBot serves ChatGPT Search, while ChatGPT-User handles user-triggered retrieval. Review these purposes and controls separately.

OAI-SearchBot

OpenAI's live search crawler for ChatGPT Search results (distinct from GPTBot).

ClaudeBot

Anthropic’s training crawler. Claude-SearchBot serves search and Claude-User handles user-triggered retrieval; these are separate crawler purposes.

PerplexityBot

Perplexity’s documented search crawler, separate from its user-requested fetcher and model-training uses. The vendor describes surfacing and linking sites; no comparative citation-quality result is established here.

Google-Extended

Google’s separate control for uses in certain Gemini systems. Googlebot controls Google Search; Google-Extended is not an additional requirement for appearing in Google Search AI features.

Schema.org JSON-LD

A way to describe page entities with structured data. Use accurate markup consistent with visible content. Google documents no special schema requirement for its Search AI features; markup alone does not establish citations or a fastest implementation path.

E-E-A-T

Experience, Expertise, Authoritativeness, Trust. Surfaced through a real named author, real credentials, and a linked entity graph (LinkedIn, Wikipedia, Crunchbase). Worth 10 points on the rubric.

Retrieval-augmented generation (RAG)

A general approach that combines retrieved information with generated text. Google describes possible query fan-out in its Search AI features. This does not establish that every engine uses one architecture, or that crawlability and HTML alone decide quotation.

Use the benchmark to identify technical and content questions to investigate. If you want a scoped report, competitor comparison and prioritized fix list, discuss the audit with me. The work can test access and clarity; it cannot promise inclusion or citations.

Prefer it done for you? Book the AI Search Visibility Audit. Scope, fees and any implementation follow-on are agreed after review.

Benchmark v1: June 23, 2026 v1.12 rescore record; historical rescore. Read the methodology and its limitations before applying these results to a current site.

This report presents a 61-site benchmark using its published rubric. If you want to understand your own technical access, content clarity and measurement gaps, book a 15-minute call to discuss fit and scope. A rubric score does not predict citations, rankings or commercial results.