The Complete Guide to LadderFox

When someone asks ChatGPT or Perplexity to recommend a tool like yours, one of three things happens: it names you, it names a competitor, or it names nobody. LadderFox measures which, repeatedly, and hands you the specific pages and threads that decide the answer.

The SEO side is here too, because Google feeds the AI answers: roughly nine in ten citations in Perplexity and Google AI Mode come from Google's top ten results. Ranking is no longer the goal, but it is still the supply line.

Who it is for

What we try not to do

Measuring AI answers is noisy: ask the same question twice and the list changes. So a score here ships with a confidence interval, a change is only called real when it lands outside the noise band, and a claim that cannot be sourced is labelled a hypothesis rather than dressed up as a finding. Where this guide describes something as automatic, a scheduler runs it. Where it does not, you press the button.

You set up a brand once. LadderFox reads your site, drafts the market you compete in, the rivals worth watching, and the questions your buyers actually ask an assistant, then shows you all of it before anything runs. You edit what is wrong and freeze it. Freezing matters: the same questions every time is what makes two runs comparable.

What a check does

Every check asks your frozen question set across the answer engines your plan includes, and records for each answer whether your brand was named without being prompted, which sources the engine cited, and which competitors appeared instead of you.

How often it runs

Scheduled checks run automatically on your plan's cadence: weekly on the free tier and Signal, every four days on Spotlight and Beacon. On top of that, each plan includes a number of checks you can start yourself, per brand, per month, so you can measure again after you have done something.

Measurement without a next step is a dashboard. Every check turns what it found into a queue of citation targets, grouped by the kind of work each one needs, with the instruction for that play next to the rows it applies to.

Marking work done, and verifying it

When you have acted on a target you mark it, and about two weeks later a verification check measures again whether the engines now cite that exact surface. The before/after comes with its noise bands, so a lift is only called likely real when it clears them.

Who mentions you

A weekly sweep tracks which sites mention your brand, with the receipts listed under the count. Mentions are verified on the page rather than inferred from a search result, and your own sister domains are excluded, because you saying your own name is not somebody else saying it.

Describe your product, or point it at your site, and five agents run in parallel (market discovery, competitor intelligence, community intelligence, technical audit and content strategy) before a synthesis step turns their output into a plan. On plans without competitor intelligence, four agents run.

What it returns

An executive summary, a plan for 90 days in three phases, eight to twelve quick wins, a twelve week content calendar, a competitive positioning statement, and the risks worth knowing about.

Recommendations carry their evidence

Every quick win cites the data points it rests on, whether that is a keyword with its measured volume, a competitor gap, a live thread or an audit finding, and those citations are checked on the server against what the run actually fetched. Anything the analysis could not tie to real data is labelled a hypothesis instead of being presented as a finding. Calendar rows whose target keyword never appeared in the fetched data are marked too.

The Opportunity Score

Arithmetic, not judgement: search volume on a log scale (0 to 30), keyword breadth (0 to 20), quick wins found (0 to 15) and share of buying intent (0 to 10), minus a competition penalty (0 to 25), clamped to a range of 20 to 95. The components are printed on the report so you can see how the number was reached. It replaced a score generated by AI that anchored near 85 for everything.

No tool asks you to type your domain again. Each opens against your configured brand, with rivals and keywords already filled in from your strategy run or from the domains the engines cite in your category. All of it stays editable per run, because which rivals to compare against is a real choice.

Research

Content

Links and authority

The audit crawls your site page by page and produces a health score out of 100 with prioritised recommendations. It checks missing and duplicate meta descriptions, broken links, missing alt text, heading structure, status code errors and schema, each graded by severity.

Core Web Vitals

Measured through Google's PageSpeed API, covering LCP, CLS, TBT and FCP, and graded against Google's own published thresholds rather than ours.

Audit history

Each audit is kept, so the health score is a line over time rather than a number in isolation. History is scoped to your account.

These two are genuinely useful and genuinely manual today. We would rather say so here than have you discover it after paying.

Position tracking

Create a campaign from your brand, with the keywords prefilled from your strategy run, and LadderFox checks where you rank in Google for each one. You get a position history chart, a visibility score weighted by click through rate, and cannibalisation detection where several keywords land on the same URL.

Reddit Monitor

A standing monitor for your subreddits and keywords. Each scan searches Reddit, scores every post for relevance to your specific product, classifies it as a tool request, a competitor complaint, a comparison or a discussion, and drafts a suggested angle for the ones worth answering. Threads you mark as replied are tracked so you can see what your comment did.

PDF reports

Any completed analysis exports as a PDF built for printing. The strategy report opens on a cover carrying the opportunity score with the arithmetic beside it, then the plan, the quick wins with the evidence each one rests on, the market and competitors, the calendar, the risks, and a closing page separating what was measured from what was synthesised from what was computed.

CSV export

Keyword lists, position tracking data, Reddit opportunities and domain analysis all export as CSV.

Sharing

Reports can be shared with a private link on paid plans, revocable whenever you like.

Alerts are deliberately quiet. After every check the run is compared against the previous one, and at most one message goes out per check. Four things qualify:

Delivery is by email to the address you sign in with, and optionally to a Slack channel via an incoming webhook. Both are off until you switch them on, and there is a test button so you can confirm the wiring rather than wondering. Every email carries an unsubscribe link.

Reports are one shared pool per plan, so a strategy run, a site audit and a gap analysis all draw from the same number, so you spend it on whatever you actually need. The free pool is a lifetime allowance rather than a monthly one. Cancel anytime from Settings; no contracts, no cancellation fees.

Language models write the summaries, plans and drafts. They do not invent the numbers: figures come from the sources above, and recommendations cite the data points they rest on.

Your analyses, keyword lists and reports are private to your account. We do not share or sell your data, and we do not use it to train models. Connections are encrypted over HTTPS and passwords are hashed with bcrypt.

Data is stored on Neon PostgreSQL. You can request full deletion by contacting support.

For the details, see our Privacy Policy and Terms of Service.