Ein Insights: why a strong Austin firm was missing from AI answers

Professional services, Austin, Texas. Pro Visibility Audit, run 27 to 29 August 2026, retested 8 September 2026.

Where it stands

Every case is scored on the same three steps. Read: can AI systems reach and read the pages that matter? Trust: do the business’s facts match everywhere they appear? Recommend: do outside sources back the business up where AI looks?

Step	Found in August	Now	Status
Read	No sitemap or robots file; a dead /examples page ChatGPT cited four times	In the fix list delivered on 29 August, with Ein Insights to make	Diagnosed
Trust	Address and phone on two chamber listings but nowhere on the firm’s own site; no Google Business Profile	In the same fix list	Diagnosed
Recommend	Absent from all four buyer questions	First pick in one retest, citing the chamber listing; still missing from the local map results	Open

This is an audit, so the result is the diagnosis. The fixes are Ein Insights’ to make.

What this proves

This is the audit case. Nothing had to be rebuilt to show the value.

In August I asked ChatGPT four questions an Ein Insights buyer would ask. Four different sets of firms came back, about 23 in all, and Ein Insights was in none of them. Asked about it by name, the model changed its mind and said it “should likely have been in the top 1 to 3”.

The audit found the reason before anyone argued about content: not a weak website, but almost no outside proof and no local business record. That is the product: a commercial diagnosis before a roadmap. What to fix first, what not to waste time fixing, and where the evidence came from.

The context

An Austin data consultancy selling enterprise-grade data science to small and medium businesses. The site is well built: ten pages, all properly linked to each other, a real page for every service, prices shown openly, and analytics and ad tracking correctly wired. None of what follows is about craft. The pattern underneath it is that the things the firm wants to be known for do not have their own web addresses.

Phase 1. Asking the questions buyers ask

The August runs asked for the best data analytics consultants in Austin, the same for small businesses, the same again with the firm’s exact tool stack named, and who can analyse Shopify, Google Analytics and QuickBooks data together.

Ein Insights was absent from all four. The local shortlist was drawn from a business index the firm is not in, and two of the four sources ChatGPT could cite about it were dead addresses.

On 8 September I asked the buying question again, in a temporary chat with memory, plugins and custom instructions off.

The map widget and the shortlist table still left Ein Insights out. But the model’s own verdict was different: it said its first call would probably be Ein Insights, and the source it pointed to was the Austin Chamber listing.

Two things keep that result honest. The chamber listing was already there in August; I did not create it. And this is one answer on one day, from an assistant that can answer the same question differently an hour later. What it shows is which outside source ChatGPT leans on when it does recommend Ein Insights, which is exactly what the audit said mattered.

Phase 2. Checking what happens to the traffic AI does send

Enumerated all eleven live addresses by hand, because there is no page list to read, then followed the stale addresses still sitting in the retrieval layer.

The /examples address returns a 404. So does /about/, which still surfaces in search results. The sitemap and robots file both existed on the previous site and are gone from this one, and two email-analytics posts, the exact topic of the webinar the firm is buying ads for, are still sitting in the internet archive.

Phase 3. Withdrawing a finding before delivery

An early draft reported that the site had no internal links at all. Before delivery I re-tested it and found the fault was mine: my pattern only matched double-quoted attributes and the site uses single quotes.

Re-checked correctly, every page links to every other page, nine to thirteen internal links each. Internal linking is fine and always was. The retraction shipped inside the delivered document with the reason attached, rather than being quietly deleted.

The results

Twelve areas scored against a services business, with the ecommerce-only checks dropped rather than scored as failures. Eleven live pages assessed by hand. Twelve ranked fixes delivered, each with the build instruction separated from the explanation, plus a plain-English version alongside the technical one.

One retest cannot prove a fix, and this page does not claim one. It shows the mechanism the audit named: ChatGPT recommended Ein Insights when it could lean on a credible outside listing, and the local results still missed it because there is no Google Business Profile. So the next commercial move is clear: complete the Google Business Profile and keep building the sources assistants use to check the business.

Evidence

---

Sources. Live tests against ChatGPT on 27, 28 and 29 August 2026; temporary-chat retest on 8 September 2026 with memory, plugins and custom instructions off. Manual enumeration of the domain. Internet archive captures. Austin and Cedar Park chamber listings. Delivered as a 12-area audit plus a plain-English version on 29 August 2026.