How AI visibility works, and how Askwords measures it.

Because an AI answer is not ten blue links. It names three to six products and stops. Ranking well helps, since it feeds the sources a model reads, but being named depends on whether your pages, and the third-party pages that mention you, answer the buyer's question directly. Askwords shows which of the two is failing: the answer text, the sources it cited, and the competitors holding the slots.

Google AI Overviews, ChatGPT, Google AI Mode and Gemini on every plan, with Perplexity added on Pro and Claude and Grok on Ultra. Askwords requests fresh answers without personal history, account memory or app personalization, so you see the default answer a new buyer would get rather than a replay of your own signed-in session.

Every answer is scored cited (1.0), mentioned (0.45) or absent (0), with a bonus when the answer names you first. Answer engines are weighted by real-world reach instead of counting equally: AI Overviews 0.35, ChatGPT 0.30, AI Mode 0.15, Gemini 0.12, Perplexity 0.08, and on Ultra Claude 0.05 and Grok 0.04. The result is then normalized across whatever engines were checked, so scores stay comparable. A failed provider call is excluded rather than counted against you.

Because they genuinely do. The same question asked twice on the same day returns overlapping brand sets only 33% to 48% of the time. Askwords treats a single flip as noise: a cell that appears to have moved is asked two more times and majority-voted, and the new result has to hold across two different days before it triggers an email. You get told when something moved, not every time an answer wobbled.

No. Pick three buyer questions and run a free check. You see every answer, the sources cited and the competing brands without signing up. A paid plan adds pinned askwords, daily checks, history, verified change notifications and the prioritized action plan.

Three levers, in the order Askwords ranks them for you: make your own pages answer the question directly, get mentioned on the third-party pages the AI already cites, and make sure AI crawlers can reach your site at all. The free check shows you the evidence behind all three. On a paid plan, each check also ends with a prioritized action plan built from that evidence: your citation gaps, your weakest answer engines and your own crawler-access problems.

Most tools in this category answer one question: are you in the answer or not. Askwords is built to close the gap rather than only report it. Sign in and each check also becomes a Steal Sheet, showing the full slate of brands every answer recommended and where you sat in it. Paid plans add a prioritized action plan drawn from that evidence, and an audit of whether AI crawlers can reach your site at all. Even the free check shows the working: the full answer text, the sources each AI cited, and which competitors took the slots, not just a number.

The real one. Askwords reads the answer chatgpt.com actually gives a buyer, because that is the one people use. The OpenAI API is a different product: it has no web grounding by default and returns a different set of brands, so reporting it as “ChatGPT tracking” overstates what was measured. Pro can track the API as its own opt-in column, deliberately given zero weight in the score, so it stays a reference point rather than a way to inflate a number.

Because most apparent changes are not real ones. When an answer looks like it moved, that cell is asked two more times and majority-voted before the new result is stored, and it still has to hold across two different days before an email goes out. A carried-over cell and two runs on the same day are explicitly not counted as independent evidence. Fewer emails, each one worth opening.

Many. The dashboard score pools every independent observation across a rolling window instead of trusting the latest run, so a cell reads “cited on 5 of the last 7 days” rather than whatever this morning happened to return. Given how much AI answers vary run to run, a single-run score is one coin flip per cell, and a tool that reports it as a fact is reporting noise.

It starts from real search demand for your business, then splits the terms into head, mid and niche bands measured against your own ceiling rather than fixed volume thresholds, so a specialist B2B tool still gets a head, a middle and a tail instead of having everything filed under “niche”. You choose which fight to pick. Most trackers hand you an empty box and ask you to invent the prompts yourself.

There is no rank in an AI answer. There are three to six named products, and you are either one of them or you are not, which is exactly why a page-one ranking and an absent brand sit together so often. Askwords also checks something rank trackers never look at: whether AI crawlers can reach you. Blocking GPTBot is a training-data policy choice, but blocking OAI-SearchBot or PerplexityBot means those engines cannot cite you at all, and plenty of sites do it by accident.