How Our Recommendations Work

The short version

A recommendation is one specific action, tied to one lever, written against your own scan — what the engines actually said, who got named instead of you, whether AI crawlers can reach your site, and which listings we could confirm you already have.

It is not a checklist with your business name substituted in. If two customers get the same recommendation, it is because the same thing was true of both of them.

We only tell you what we verified

Every check we run has three outcomes, not two: found, absent, and unknown. Absent means we looked and it is not there. Unknown means the check itself did not finish — an API error, a timeout, an expired key on our side.

Anything unknown is left out of your recommendations entirely. We do not report it to you as a gap, and we do not mention it at all.

This rule exists because of a real failure on an earlier tool of ours. A security check broke because of outdated software on the machine running the check — and the report told the customer their website was using an obsolete security protocol. Our failure, stated as their fact, in a product whose entire claim is that it went and looked. It takes exactly one of those to lose the room, so the rule is now absolute.

The evidence standard

A recommendation has to be documented by the engine, or measured by us. Not merely plausible. A great deal of AI-visibility advice currently sold is plausible and unverified, and it is impossible to tell apart from the real thing without doing the work.

The clearest example is schema markup, which most of the industry still recommends and we do not:

Given that, telling you to add markup so that AI engines find you would be an unsupported claim stated as a fact. So we spend the recommendation on something with better evidence behind it.

Just as importantly, we never tell you to remove markup you already have. It continues to earn its keep through Google rich results, which is a separate benefit from AI search. It is a lower priority than it once was — not a mistake to be undone.

Which directory matters depends on the engine

“Get listed in directories” is standard advice and it is too vague to be useful. This one is not a vendor statement — it is our own measurement, taken across 144 real AI responses on 9 August 2026:

So directory work pays off on Google and Gemini, and website quality is what moves ChatGPT — which is the largest engine. This is why a recommendation tells you which engines it is expected to affect, rather than implying every action helps everywhere.

One problem outranks everything else

If AI search crawlers cannot reach your website, nothing else you do can work. OpenAI states that sites disallowing its search crawler do not surface in ChatGPT search answers, and ChatGPT sources local answers overwhelmingly from businesses' own websites.

When we detect that, it becomes your top recommendation and everything else waits behind it. It is also frequently not your fault and not a code change: several security providers began blocking AI crawlers by default on new sites in July 2025, so the fix is often a setting someone can change in an afternoon.

This is the one case where the free plan sees a technical recommendation. Telling someone everything except the reason they are invisible would be worse than telling them nothing.

Fewer, and stronger

We ask for up to five recommendations per scan on Starter and seven on Pro — and returning fewer is explicitly allowed. Two strong recommendations beat seven where five are padding.

Every lever already suggested to you is off limits for future scans, so a business that has been scanned for months genuinely runs out of new things worth doing. When that happens you get a short list, or none. That is the honest answer, and a recommendation you can dismiss as irrelevant costs more trust than a short list does.

Why your list is a mix, not five red badges

Your recommendations are deliberately a spread across priorities rather than every high-priority item at once. That is a choice, and it is worth knowing the reason.

Some of the most valuable actions are also the slowest. Creating a Google Business Profile means waiting on a verification postcard. Earning a mention on an industry site depends on someone else's timeline. If your whole list were things you cannot finish this month, you would be stuck — so it also carries quicker, lower-priority work you can complete while the slow items are in progress.

This means a new high-priority recommendation can appear after you have cleared the ones you had. That is the list refilling as you complete things, not something we held back. Priority describes how much an action is worth, and the mix exists so you always have something you can actually do today.

Why new recommendations stop appearing

Once you have twelve open recommendations, indextr stops generating more until you work through some of them. Scans keep running and your score keeps updating — only the list pauses.

The reason is that the value of a to-do list collapses as it lengthens. Before this limit existed, someone who signed up at a trade show on a Wednesday could open the app on Friday to roughly forty-five items, which is not a plan — it is a wall.

Note what we did not do: we never hide recommendations that exist. Truncating a displayed list can bury a cheap, easy win behind a row of harder ones, and you would never learn it was there. Declining to generate while a backlog is outstanding hides nothing.

What each recommendation tells you

Every recommendation carries three things beyond the action itself.

Competitive position is scored but is deliberately not offered as a target, because it is not a lever. It moves as a consequence of the other four shifting relative to everyone else. Tagging a recommendation with it would be true of nearly all of them and actionable on none.

Marking them done

You tick recommendations off as you complete them, and completed items are fed back in so the same advice is never given twice.

We also tick some off automatically when a later scan proves the work was done — if you create a Google Business Profile, the recommendation to create one closes itself. But that runs under a deliberately narrow rule, because the two mistakes are not equal. A recommendation left open by mistake is visible and you can dismiss it. One completed by mistake disappears, and because completed items are never suggested again, it is suppressed permanently. So when in doubt, we leave it open for you to judge.

What each plan gets

The paid plans differ in how much is measured, not how much is explained. Pro scans five engines instead of three and thirty search phrases instead of fifteen — but Starter gets exactly the same depth of instruction.

That was a deliberate decision and it cost us an obvious upsell. This product exists so that a small business owner who may or may not touch their own website can make these changes without paying an agency a great deal of money. If that is what it is for, the how is the product — and putting it behind the higher tier would make the entry plan not the product.