AnswerShare Proposal for Texas First Rentals

Frequently asked questions

Short answers to what clients ask us most.

Is this cloaking?

Different formatting is not, by itself, deceptive cloaking, and this deployment is built to avoid it. Google Search Central defines cloaking this way: “Cloaking refers to the practice of presenting different content to users and search engines with the intent to manipulate search rankings and mislead users.” See Spam policies for Google Web Search. The definition turns on showing search engines something different from what users see, in order to manipulate rankings.

Who receives what:

• People: your normal website, unchanged.
• Every declared search crawler, including Googlebot and Bingbot: your normal website, unchanged, because search engines require their crawlers to see what people see. They are never sent the machine layer.
• Every declared AI crawler, for example the training crawlers GPTBot and ClaudeBot, the AI search crawlers OAI-SearchBot, Claude-SearchBot and PerplexityBot, and the user-triggered fetchers ChatGPT-User, Claude-User and Perplexity-User: a machine-readable representation derived from the same approved facts as your website.
• Nothing is hidden. Routing is by the crawler’s declared User-Agent, not by IP address, so anyone can request a page as an AI crawler and see exactly what it receives. Your site’s llms.txt and llms-full.txt are also published openly at public addresses.
• The machine layer restates facts your site already publishes and your team has approved. It is not a separate set of claims written to influence search rankings.

Same text as in the Presentation tab, section "Is This Cloaking?": https://proposals.answershare.com/efa83680e8538e264aa420f1d791d6cf/presentation

• The check is a URL-by-URL parity and canonical audit (including deletions, redirects, noindex and stale-cache cases), with side-by-side captures for people, verified search crawlers and AI agents, and a stop-and-rollback threshold if Search Console or Bing Webmaster Tools show a crawl or indexing regression. Google’s definition describes deception; it is not Google’s approval of this deployment.
• Anyone can send a crawler’s name in a request, and routing follows that declared name by design. An impersonator receives only the machine layer: your own approved public facts, which are already on your website. Faking a crawler gains them nothing. Crawl counts are reported as self-identified requests (the declared User-Agent), not as verified provider traffic.

Scope of this answer: it applies Google’s published cloaking definition to search crawlers and search rankings.

Google AI Overviews and AI Mode use Google’s own search index, which reads your normal website, so they are improved by fixes to the human site, not by the machine layer. This answer does not speak for how any AI company evaluates content, and AI companies decide for themselves what to retrieve and cite. The same reasoning holds for the human-versus-machine format difference as for a mobile layout, an RSS feed or a plain-text version of a page: the same information, for a different reader.

ETHICS & TRANSPARENCY

Q. Is this cloaking?

A. DIFFERENT FORMATTING IS NOT, BY ITSELF, DECEPTIVE CLOAKING.Google's own definition of cloaking:

> “Cloaking refers to the practice of presenting different content to users and search engines with the intent to manipulate search rankings and mislead users.”

> — Google Search Central

Googlebot and Bingbot receive the human site. AI crawlers receive a machine layer derived from the same facts, restated for AI readers.

SOURCE: Spam Policies for Google Web Searchdevelopers.google.com/search/docs/essentials/spam-policies#cloaking

AnswerShare

Can’t we just use a markdown translation, like Cloudflare’s Markdown for Agents?

You can, and the two do different jobs. Markdown for Agents converts the pages you already have into markdown when an AI agent asks for that format with an Accept: text/markdown request header; every other request gets the normal page. It is available on Cloudflare’s Pro, Business and Enterprise plans. That makes each page easier for an AI to read, but it can only pass on what that page already says.

AnswerShare does a different job. It gathers the facts spread across your whole site and your public record (your services, locations, credentials, and ratings with their sources, plus an FAQ your team has reviewed) into one consistent briefing that AI assistants can read and cite. It keeps that briefing current when you publish, and measures how AI answers about your business change over time.

What is the contractual commitment?

When you approve, we implement our entire solution. We need you to answer a few questions. That takes about ten minutes. Your term commitment is 90 days. Thereafter, you can stop anytime: tell us, and we agree the timing together so your service is never interrupted.

How does this affect the way AI assistants describe our brand?

AI answers hedge or get details wrong when they cannot find clear facts. AnswerShare gives AI assistants your facts in a form they can read and cite: your services, locations, credentials and ratings, each tied to its source. We measure how AI answers about you change over time, but no one can guarantee what an AI will say. Press coverage and third-party reviews stay with your PR team.

Do we need to change our CMS or rebuild the website?

No. Your site, design, CMS and publishing workflow stay as they are. Setup is one DNS or Cloudflare change, made with your DNS administrator on a short call, plus a publish webhook where your CMS can send one, so your updates reach AI crawlers quickly.

How does this fit with our SEO work?

It sits alongside it. Your SEO team or agency keeps working on Google rankings, keywords and site performance, and nothing we do changes what Google’s search crawler sees. AnswerShare works on how AI assistants such as ChatGPT, Claude and Perplexity find, read and cite your business.

What do we need from our team?

A little setup time, mostly in the first weeks. You answer a few questions on your onboarding page (about ten minutes). Later, your DNS administrator makes one DNS or Cloudflare change on a short call after the pre-cutover tests, and you review the FAQ we prepare from your public materials when it’s ready. If you use them, you give us read-only access to your analytics and Search Console, and any SEO keyword or page-planning exports you have. No custom code is required.

The pilot asks more of your team than setup does: it also approves the content eligible for the machine layer, configures and maintains the outbound publishing notification, and takes part in the acceptance tests in the Technical tab.

To keep the layer current automatically, your build or deploy step sends one signed change notification (we supply the specification); otherwise we agree a manual update process in writing.

Separately, corrections and new pages on your own website, and counsel review of any legal or regulatory text, are your team’s work and are not part of the AnswerShare investment.

How the scores work

What is the GEO score, and how is it scored?

GEO (Generative Engine Optimization) is a score from 0 to 100 for how ready a website is for AI assistants to find, read and cite. It grades the website. It does not grade what any AI currently says about the brand. GEO is AnswerShare's internal readiness indicator for whether the machine layer is AI-optimized; we do not measure success or failure by it, and success is judged on the pilot scorecard outcomes.

We collect one evidence record from the live site: the home page, robots.txt, the sitemap, llms.txt and llms-full.txt, and response timings. Five AI engines then grade that same record against the same fixed rubric: Perplexity, OpenAI, Gemini, Claude and Grok. Some of these engines can also search the web while they grade.

The rubric has nine sections worth 100 points in total:

• AI crawler accessibility: 10 points
• AI context layer: 16 points
• Entity authority: 14 points
• Structured data: 10 points
• Content answerability: 12 points
• Citation utility: 14 points
• Retrieval efficiency and freshness: 12 points
• AI retrieval performance: 8 points
• Crawl infrastructure: 4 points

Missing pieces lower the score instead of failing the audit. Two caps apply. A score above 85 needs clean AI files, sitemap, robots.txt, schema, canonical routing and basic performance. A score above 90 also needs strong citation utility, efficient retrieval, reliable freshness signals and excellent performance.

The headline is the median of the engines’ scores, not the average. If the five engines disagree by more than 10 points, the one engine furthest from the median is left out and the median of the other four is published. Nothing is left out when fewer than five engines scored, or when two engines are equally far from the median. A headline needs valid scores from at least 4 of the 5 engines. With fewer, it shows no score instead of a low-confidence one.

The bands are Invisible (0 to 20), Faint (21 to 40), Emerging (41 to 60), Established (61 to 80), Leaders (81 to 90) and Gold Standard (91 to 100). Our house target is 85 or higher.

What is the SEN score, and how is it scored?

SEN (sentiment) is a score from −100 to +100 for the opinion AI engines express about a company: its reputation, trustworthiness and standing. −100 is very negative, 0 is neutral and +100 is very positive. It is separate from GEO, so a technically strong site can have a modest SEN, and the reverse.

Each of the five engines gives one sentiment value when it grades the site. The published SEN is the median of the values that came back. A missing value is left out. It is never counted as zero, and SEN shows as not measured when no engine returned one.

We read the number simply: positive is favorable, zero is neutral, and a negative value is worth a closer look. SEN has no other named bands.

Proposals also show an AI Sentiment star rating, from 1 to 5, for the brands compared. We ask each engine about the brand directly and read every answer for the stance it expresses. The share of measured answers that express trust sets the stars on a fixed scale: 5 stars at 90% or more, 4 stars from 75% up to 90%, 3 stars from 60% up to 75%, 2 stars from 40% up to 60%, and 1 star below 40%. The scale is absolute. It is not a curve against competitors.

The rating drops to 1 star when an engine’s own answer cites a failing Better Business Bureau grade, advises against the brand, alleges fraud or a scam, or reports an active legal or regulatory action. Each one is recorded with the sentence the engine wrote. If no answer could be measured, no stars are shown.

What are these scores useful for?

• GEO shows where AI engines struggle with the site. The nine sections, and the biggest gaps the engines name, point to the fixes that matter most.
• SEN shows whether engines describe the company favorably, neutrally or unfavorably. A weak or negative reading is a reason to find out why before a buyer meets it in an AI answer.
• Measured before and after the work, the same rubric shows what changed. That helps decide where to spend next.

They describe your own site and your own brand. They are not a ranking against competitors, and they are not a forecast of traffic or sales.

What are the limitations of these scores?

• A score is a dated measurement, not a forecast. It says what we measured on that day. It does not predict what AI will say later.
• Answers vary from run to run. AI engines do not give identical answers every time, so a re-run of the same site can score differently. Using the median of the engines, with an outlier engine left out, reduces that variation. It does not remove it.
• Some engines search the web while they grade. A change on the web between two runs can move a score.
• The graders are AI models applying a fixed rubric. A score is their combined judgment, not a lab test or a human audit.
• Scores are comparable only under the same method. If the scoring prompt version, the set of engines or an engine’s model changes, scores before and after that change are not directly comparable. We record these versions with every run.
• Measures built on buyer questions, such as the AI Sentiment stars, use a chosen sample of questions, not every question a buyer could ask. If the question set is edited, readings before and after are not like-for-like. An answer that could not be measured is left out, not counted as zero.
• SEN is the opinion the engines express. It is not a review-site rating, and it does not read customer reviews directly.
• These scores measure what an AI says. They do not control it.

Next: Part 5 of 7 — AI Prompt Results https://proposals.answershare.com/efa83680e8538e264aa420f1d791d6cf/results

End of the FAQ tab of the Texas First Rentals proposal (12 sections).