Key Takeaways
- Expect the work and the result to run one month apart. Publishing is fast and being absorbed is slow. Almost nothing you do in the first thirty days shows up inside the first thirty days.
- Front-load the boring page work. Services, city, prices, method, and results in plain text. It is the least exciting month and everything after it depends on it being done.
- Judge month two by movement, not by winning. The right question at day 60 is whether any answers changed at all, not whether you now top the list. Direction beats position this early.
- Re-run the identical question set at day 90. Same wording, same assistant, same method. A comparison against a set you edited along the way tells you nothing you can act on.
- Treat 90 days as the first readable signal, not the finish. It is long enough to know whether the work is landing and short enough that the slowest half of the problem has barely begun to answer.
Ninety days is the number everyone in this market quotes, and almost nobody says what happens inside it. So it lands as a promise instead of a plan.
The honest version is less dramatic and more useful. Most of the work happens in the first sixty days, most of the visible change happens in the last thirty, and the two halves are offset from each other by about a month.
Here is what fills each block of thirty days, what you can fairly expect to see by the end of it, and what ninety days cannot deliver no matter who is doing the work.
What does 90 days of AI visibility work actually look like?
One month making your own facts readable, one month getting named on pages you do not own, and one month filling gaps and re-measuring. The visible change trails the work that caused it by roughly a month.
That lag is the single most useful thing to understand before you start. Publishing happens the day you press publish. Being read, quoted, and recalled happens on somebody else's schedule, and no amount of urgency moves it.
Which means the honest promise for ninety days is a readable signal and a fair comparison, not a finished result. A provider offering more than that is selling the part nobody controls, as covered in what AI visibility services cannot promise.
Month one: the part nobody sees
The first thirty days go on the baseline and on your own pages. It produces almost no visible change, and skipping it makes everything afterwards unmeasurable.
The baseline comes first: the 25 questions run and written down, with who gets named where you do not. Everything at day 90 is a comparison against this, so a baseline taken casually is a baseline you cannot use.
Then the unglamorous page work. Services named plainly, the city written down, prices or a range stated, the method described as steps, and client results moved out of image files and testimonial sliders into text a stranger could quote. Most businesses fail on this alone, which is why it comes first.
- The 25 questions run and written down
- Services, city, and prices put into plain text
- The method described as steps on a real page
- Results rewritten out of images and sliders
- Association and niche directory listings corrected
- Client results published as readable case studies
- Reviews asked for in a form that quotes well
- Two or three third-party pages that name you
- The gaps the first run exposed, filled
- Comparison and pricing questions answered on the site
- The same 25 questions re-run, same wording
- A before and after readout
The work of month one has been done and not yet read by anything. This empty block is the most common reason people quit in week three.
- Live-retrieval assistants can quote the new pages
- Your own name returns better, more specific answers
- Some direct discovery questions start naming you
- Enough questions have moved to be a trend, not noise
- Which competitors hold their positions becomes clear
- The remaining gaps are the ones needing more time
The right-hand column is pushed down one block on purpose. Each row of the figure is one calendar month, so what shows in a month is caused by the work of the month above it. The blocks describe the shape of a typical ninety days and are illustrative rather than measured.
Month two: evidence you do not own
The second thirty days go on getting named somewhere other than your own site, because that is what the slower half of the problem runs on.
The list is short and dull: association and niche directory listings with your facts correct and consistent, client results published as case studies a stranger could read, reviews asked for in a way that produces quotable sentences, and two or three genuine third-party pages that name you.
Month two is also when the first movement appears, from month one's work. That timing catches people out. It looks like the directory listings worked instantly, and what actually happened is that the page work finally got read.
| Checkpoint | Fair to expect | Not yet | Why |
|---|---|---|---|
| Day 30 | A baseline, and your own pages rewritten | Any change in who gets named | Nothing has been re-measured, and little has been re-read |
| Day 60 | Movement on live-retrieval assistants and on branded questions | Movement on questions answered from training | Recall is built from repetition over months, not weeks |
| Day 90 | A same-wording re-run, and a readable trend | A settled position you can hold | The slowest half of the problem has only just begun to answer |
| Day 180 | Whether the trend held or stalled | A guarantee of any specific placement | Nobody controls what an assistant says, including us |
Month three: re-measure and read the trend
The last thirty days close the gaps the baseline exposed, then re-run the identical 25 questions and compare. Identical is the operative word.
Same wording, same assistant, same order, written down the same way. A comparison against a question set that was quietly improved along the way is not a comparison, and it is the most common way this work gets misreported.
Read the result as a direction. Two or three more questions naming you, out of twenty-five, is a real and normal ninety-day outcome for a business that started invisible. It is also small enough that only the written baseline proves it happened, which is the case for taking the measurement seriously in week one. The signals worth reading are covered in how to tell if AI visibility is improving.
What 90 days cannot deliver
It cannot deliver a guaranteed position, because nobody controls the answer. Google says plainly in its guidance on AI features in Search that there is no separate system to optimize against, and OpenAI's description of what ChatGPT is describes a model generating from what it absorbed, not a ranked list anyone can edit.
It also cannot compress absorption. You can do every piece of the work in the first week and the part that depends on your name being repeated across independent sources over time still takes the time it takes. Paying more buys the work done faster, not the world reading it faster.
What it can deliver is a written baseline, the fixable half genuinely fixed, a handful of third-party pages that did not exist before, and an honest comparison at the end. Whether that is worth paying for is a separate question, worked through in is a paid AI visibility audit worth it.
Our own published bound is deliberately modest for the same reason: one net-new citation in 90 days, or 90 more days free, and a full refund if there is still nothing at day 180. A number anyone can check beats a promise nobody can keep.
If you want the baseline half of this done for you first, the free check runs 25 real buyer questions from your niche through ChatGPT at $0 and emails a plain-English readout within the hour. It runs as soon as you ask for it.
Questions we hear the most
What does 90 days of AI visibility work actually look like?
Roughly one month of putting your own facts into readable text, one month of getting named on pages you do not own, and one month of filling gaps and re-measuring. Visible change tends to run about a month behind the work that caused it.
How long before AI starts recommending me?
Live-retrieval assistants can reflect new pages within days or weeks. Assistants answering from training change over months, so a realistic first readable signal is around day 60 and a fair comparison is at day 90.
Why does nothing seem to happen in the first month?
Because the first month is spent making facts readable, and nothing has been re-measured yet. The work is real and the measurement that would show it comes later, which is why the baseline is taken on day one.
Is faster possible if I pay more?
Money can compress the work, not the absorption. You can publish everything in a week, and the part that depends on a name being repeated across independent sources over time still takes the time it takes.
What counts as a real result at day 90?
A measurable change in how often you are named across the same 25 questions, run identically. Anything based on a different question set, a different assistant, or a remembered impression is not a result.
What does the $500 audit include?
It runs all 125 checks across five AI tools and lands in your inbox by this time tomorrow. It is the full baseline version of the free check, and the $500 is credited if you go further.
Does AnswerHalo guarantee results in 90 days?
The published guarantee is one net-new citation in 90 days or 90 more days free, and a full refund if there is still nothing at day 180. No honest provider guarantees a position or a ranking, because nobody controls the answer.
Can I run the free check today?
Yes. It runs as soon as you ask for it, and the readout lands within the hour between 7am and 5pm Central. Taking your own baseline by writing down who gets named in your buyers questions costs nothing and can happen today.
Your buyers are already asking. Find out what AI tells them.
The free check asks ChatGPT 25 real buyer questions about your niche and sends you the report within the hour, at $0. If it shows you are already getting named everywhere, we will say so in plain words.
Get my free check →