How Logres checks AI visibility without making promises

Anyone can ask an AI assistant one question and take a screenshot. A useful business check needs more discipline: consistent questions, saved full answers, clear scoring rules, and honest limits.

Here is how Logres approaches that work.

Use the same buyer questions

Each check starts with a fixed, versioned group of natural questions that buyers ask when choosing or comparing businesses.

Logres saves every full answer with its date and any source links shown. Keeping the questions and check version steady makes a later comparison more useful. If the setup changes too much, Logres does not present the difference as a trend.

Keep four results separate

Logres reports four results on separate 0–100 scales:

• Visibility: How often and how prominently your business appeared in saved recommendation answers.
• Authority: How often your own public pages appeared among the source links shown in decision answers.
• Accuracy: How closely saved descriptions matched facts your business approved, when enough information was available.
• Influence: Whether the answer gave buyers useful reasons to consider your business.

These results are never averaged into one overall number. A business can appear often but be described incorrectly. It can also be described well while its own website is rarely shown as a source. One average would hide the problem you need to fix.

Leave unsupported results unavailable

Some scores require specific information. Accuracy, for example, needs an approved fact sheet. Without one, there is no fair way to judge whether a description is correct.

Logres marks that result unavailable and explains what is missing. It does not estimate a score to fill an empty dashboard cell.

Report change without claiming cause

A score or answer may differ in a later check. That difference is worth reviewing, but it does not prove why the change happened.

Models, assistant products, outside sources, and public pages can all change. Logres shows what went live on the website and what assistants said later while keeping those steps separate.

What you gain

You receive a dated first check you can inspect, a prioritized list of public gaps, and a way to check again under compatible rules. That is less dramatic than a guarantee, but it gives you something real to act on.

Common questions

Why not combine the four results?

Because each identifies a different problem. One overall score would hide whether you are missing, poorly described, rarely cited, or weakly presented.

What does unavailable mean?

It means the information needed for a fair score was missing. The report explains what is needed instead of guessing.

If a later score changes, did the work cause it?

Not necessarily. Logres can show the timing and the differences, but a before-and-after sequence alone does not prove cause.

How a finding is supported

Buyer question→Saved answer→Source + date→What it means→Clear finding

Research notes

Sources worth reading alongside this note

These primary or first-party sources give more context for the ideas in this article. They do not prove anything about a Logres client or individual result.

1. 01

   Artificial Intelligence Risk Management Framework 1.0

   NIST

   A practical vocabulary for governing, mapping, measuring, and managing AI systems.

2. 02

   Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks

   Lewis et al., arXiv

   Foundational research on retrieving explicit sources alongside model memory.

3. 03

   Introduction to structured data markup

   Google Search Central

   Google's documentation on making entities and page meaning machine-readable.

This guide is for information only. Logres checks AI answers for an agreed list of questions on a specific date. It cannot promise that an assistant will name, link to, or recommend a business later, and it does not predict rankings, traffic, leads, or revenue.