The Cost of Doing Nothing

What the audit found in July

A working real-estate site — real listings, a real agent, a business doing fine. None of the following was visible to anyone using it. All of it was visible to a machine deciding whether to show it to anyone else.

1. A noindex tag
   The site was telling search engines not to list it. Not ranking badly — absent. Every other improvement is worth nothing while that one line sits in the head of the page, and nobody had noticed it was there.
2. Split www and non-www
   The two addresses did not resolve to the same URL, so search engines saw two different sites. Whatever authority the site earned was divided between them instead of accumulating in one place.
3. No XML sitemap
   New pages had no reliable path into the index. On a listing-heavy site that is the whole problem: a listing indexed late is a listing indexed after it sells.
4. A robots.txt error
   The crawl rules could not be read, so what a crawler was allowed to fetch was left to guesswork — on a site already carrying a tag telling it to leave.
5. No meta description
   Search engines wrote the snippet themselves from whatever text they found first. The one sentence a searcher reads before deciding whether to click was being written by a machine that had never seen the business.
6. Six H1 tags on one page
   There should be exactly one. Six leaves the subject of the page genuinely ambiguous to anything trying to work out what it is about.
7. Images with no alt text
   Invisible to image search and to screen readers. On a property site the photographs are the product, and they were unreadable to the two systems most likely to surface them.
8. No Open Graph tags
   Every share on Facebook or LinkedIn rendered bare — no image, no description, no reason to click.

What the re-check found 25 days later

The site was re-checked on 21 August 2026 against the on-page visibility signals. One July problem is fixed: the social preview tags now pass. Two are confirmed still failing by both checks — no meta description, no sitemap. And the one that decides whether the site can be found at all is untouched: it is still telling search engines not to index it.

• Still no meta description — Failing in July. Failing in August. Confirmed by both checks.
• Still no sitemap — Failing in July. Failing in August.
• No FAQ schema — The single heaviest signal for getting quoted by an AI engine, and the site has none.
• No Speakable markup — Nothing on the page is flagged as answer-ready, so an engine looking for a passage to lift has to guess.
• No location mentioned — The pages do not name the area they serve, which is the signal that decides whether an engine recommends the business locally at all.
• No llms.txt — Nothing tells an AI crawler what to read first.
• No neighborhood or location pages — Nothing targets the "in this area" questions buyers actually ask before they call anyone.

Seven signals missing, and a tag instructing search engines not to list the page. Every one of these is independently checkable — the same signals the free score check reads on any site, including yours.

The pattern is worth sitting with: the presentable thing got fixed. The one that costs everything did not. That is what doing nothing usually looks like from the outside — not neglect, just never getting to the hard one.

Being straight about the limits of this comparison: the two checks are different tools. The July pass was a website health audit; the August one reads on-page visibility signals. Four July findings — the domain redirect, the number of H1 tags, image alt text and the robots.txt error — are simply not things the August check looks at, so this page does not claim anything about them either way.

What was built, and never turned on

The rebuild was completed on staging back in July. It fixed everything fixable away from the live domain: the meta description, a single H1, alt text on every image, the sitemap, the Open Graph tags, and the noindex tag removed. The two remaining items — the www redirect and robots.txt — were configured in the build and validate the moment DNS points at it.

The domain was never pointed at it. There was no invoice and no contract; the work was already done and waiting. It is still waiting, and the live site is still serving the problems above.

Why broken sites stay broken

It is almost never the money. It is the fear that changing something will break something that already works — that a switchover goes wrong, that the phone stops ringing, that a problem you cannot see becomes a problem you can.

The trouble is how much weight "already works" is carrying in that sentence. A site with a noindex tag is not working. It is running. Those are different things, and only one of them shows up in the analytics you are afraid of disturbing.

Doing nothing feels like the cautious option because its cost never arrives as an event. Nothing breaks on the day you decide. The bill turns up somewhere else — in the listing that indexed after it sold, the search that surfaced another agent, the shared link that rendered blank, the buyer who asked an AI who to call and heard a name that was not yours. None of that appears in a dashboard as a loss. It appears as an ordinary quiet month.

The honest part

We cannot tell you what this cost, and we are not going to invent a number. Nobody can measure a search that never happened. What is measurable is on this page: the failures found in July, the ones still failing in August, and a fix that has existed and gone unused the whole time.

That is the argument, and it is the same one on the case study where somebody said yes: a 12-year shop that went from zero AI citations to nine in 45 days. Same method, opposite decision. The method is published in full either way.

Questions

Find out what your site is doing while you think about it.

Run the Free AI Score Check →

Or get the named list and the fix order: Site Inspection — $99