What's new in 2026

AI Mode tracking + GA4 Insights

The full platform

Every SEO tool you need — built for the AI search era

SearchChamp bundles keyword research, rank tracking, site audit, AI content, AI search visibility, and agency reporting into one platform. Every problem our audit finds arrives with the move that fixes it — each gap routed to a fix, staged with a before-and-after preview, and reversible. Transparent pricing, no add-on stacking.

Start free trial See pricing

The platform

One platform for search that answers back.

ACT IWhat AI actually sees

01Site Auditor

Google renders your JavaScript. GPTBot doesn’t.

Every page is fetched twice — once the way a browser sees it, once the way a non-rendering AI crawler does — and the two results are diffed. A heading, a schema block or a price that only exists after JavaScript runs gets flagged as invisible to the crawlers deciding whether to cite you. Your site looks finished in your own browser and half-empty to GPTBot — a difference that only appears when the two fetches are diffed against each other.

Raw HTML against rendered DOM, page by page, with the missing headings, price tokens and JSON-LD types named individually.
An AI-readiness score whose recipe is published rather than proprietary — Access 40 · Structure 25 · Extractability 25 · Credibility 10, out of 100. Alongside it, a separate per-crawler readout scores the same page for Googlebot, GPTBot, ClaudeBot, PerplexityBot and GeminiBot against each one’s own extraction rules.
Findings are ranked by estimated traffic lift computed from your own Search Console numbers, not by severity — a page that fell from one click to zero over the trailing 28 days outranks a missing alt tag. Work top-down and stop when it stops mattering.

Parity — two fetches, one page

GPTBOT (RAW HTML)

MISSING

BROWSER (RENDERED)

H2JS-ONLY

SCHEMAJS-ONLY

PRICEJS-ONLY

Directional signal. AI crawlers that do not execute JavaScript only read the raw HTML column.

SORTED BY ESTIMATED LIFT

/pricing — price renders only after JSest. +41 clicks

FAQ schema absent in raw HTMLest. +27 clicks

Heading outline invisible in the raw fetchest. +12 clicks

Missing alt text on product imagesest. +2 clicks

02AI Crawler Analytics

858 hits claimed to be GPTBot. 162 were telling the truth.

Every tool that counts “AI bot traffic” is counting user-agent strings — text any script can type — because crawlers never execute JavaScript and a server log records only the claim. We match each hit against its real source IP and the crawler operator’s own published address ranges. On one real account in one recent week, 696 of the hits claiming GPTBot came from somewhere else entirely, and every single hit claiming ClaudeBot failed the check.

Five verdicts instead of a yes/no — Verified, Spoofed, Unverifiable, Pending check, Check degraded — because “we can’t prove this” is three different things and one of them is our outage, not yours.
It will not accuse on thin evidence: no cached ranges means Pending, a stale snapshot means Check degraded. Spoofed is only ever asserted against a fresh one.
Then it shows what your own robots.txt actually says to each of them, in three states — because Not addressed is not the same as Allowed and we never render it that way. robots.txt is advisory, not enforcement; well-behaved crawlers honour it, and we won’t sell you a switch that pretends otherwise.

Claim vs published IP range

858 hits claiming GPTBot

Verified · 162Spoofed · 696

78 hits claiming ClaudeBot

Spoofed · 78 of 78

FROM THE OPERATOR’S PUBLISHED IP RANGES

VerifiedSpoofedUnverifiablePending checkCheck degraded

03AI Visibility Tracker

Track your brand inside AI answers — every number ships its sample size.

Run your buyers’ real prompts against the answer engines and record who got cited and where. Your top prompts checked daily, the rest weekly — you choose the split. The difference from every other AI-visibility dashboard is what this one refuses to draw: a share-of-voice figure never appears without the sample it came from, and a move smaller than its own confidence interval reads “within noise” rather than a trend.

Engines are named, never totalled: ChatGPT, Google AI Overviews, Google AI Mode, Gemini and Microsoft Copilot.
Under thirty sampled runs a figure is labelled “directional (low sample)” and is denied both an interval and a coloured arrow — the method is a 95% Wilson interval, chosen because the normal approximation under-covers exactly the sample sizes a weekly-refreshed prompt lives in.
Prompt value renders as an em-dash when we can’t anchor it to a real keyword — never $0 — and roll-up totals state their own coverage rather than summing an unknown to zero.

Share of voice — with its sample

18.4%±▲ 2.1

SAMPLE · n=259n=12 · directional (low sample)

03BAI Visibility · Action Layer

Every gap arrives with the move that closes it

Measurement stops at a chart. Each answer you are missing is routed to a specific, evidence-backed move — and each move starts in one click from the row that found it.

LANE 01 · THE PAGE EXISTS

Fix an existing page

A page of yours is already cited, but the answer built from it skips what the buyer asked. The gap arrives with that page, the crawl evidence behind it, and the sub-questions the answer missed — and when nothing on your site is a confident match, it says so rather than nominating the least-bad page.

Fix an existing page

LANE 02 · NO PAGE CAN CARRY IT

Write new content

Nothing on your domain answers the question at all, so the answer was assembled from someone else's material. The gap becomes a brief carrying the question, the rival pages it was built from instead, and an honest “unknown” where reach cannot be estimated.

Generate content brief

LANE 03 · THE PAGE HOLDS, AUTHORITY DOESN’T

Earn a mention

A third-party source is cited talking about a rival and never mentioning you. It is listed as a pitch target only when it was actually observed citing one of your tracked questions — never a generated prospect list — and it lands on the outreach board as a draft you approve before anything is sent.

Start outreach

VERIFICATION

Re-checked when the AI re-reads the page. The re-check is triggered by the crawl rather than by a calendar: when a verified AI crawler fetches the changed page again, that answer is re-run and the before and after sit side by side — including the runs where nothing moved. Read it as directional, not conclusive. It is a single sample of an answer that drifts day to day, so it reports what changed on each side of your edit and never claims your edit is the reason.

re-run on next scheduled checkoutcome recorded either way

What your plan governs

Action routing is not tiered — wherever a question is tracked, its routed lanes come with it. What the plan sets is how much you measure.

Read live from the billing model the scheduler enforces, not from a hand-written table.
Plan	Prompts tracked	Engines per prompt	Refresh
Starter	30	5	weekly
Business	100	5	weekly
Agency	250	5	weekly

04How we handle what we don’t know

Not checked is never shown as clean.

A number we didn’t measure is never rendered as a number. A score that was never computed shows an em-dash and says why. A volatility reading below its sample threshold says “collecting”, not “low”. A keyword with no signal says “Not yet scored”, not 0. A crawler hit we couldn’t check says “Pending check”, not “Spoofed”. A run that half-finished says “4 of 6 phases finished” and never a full bar. This is the most expensive thing we build, because it has to hold in every query, every score and every empty state.

Absence has a vocabulary, and the words are load-bearing: Not yet scored No data Not ranking Not addressed Found nothing Didn’t complete Collecting Pending enrichment Couldn’t read Not run yet
When a dimension wasn’t measured the formula changes rather than scoring it zero — the site health score drops its performance term and re-weights the remaining three, byte-identical to the score before that dimension existed.
Unknown behaves as failure everywhere it costs us: an unresolved plan is treated as unpaid so no paid source arms itself, an unparseable spend ceiling reads as already spent, and a thin evidence corpus makes the outbound-email guard stricter, never looser.

SELF-AUDIT

We audit this site against the running product and delete the claims that fail. A page that said no agent or script is installed on your site — false, capture needs a plugin or an edge worker — is gone. So is one that promised a rival-page report nothing shipped.

The dashboard, told honestly

what a normal dashboard showsALL CLEAR ✓

AI-READINESS

74/100

AI-READINESS0

VERIFIED HITS

162

VERIFIED HITS0

CITATIONS · WK

CITATIONS · WK0

VOLATILITY

3.1

typical night lands on 3

VOLATILITY0

PERFORMANCE

—

not probed this run

PERFORMANCE0

Not probed this run — the health score drops this term and re-weights the remaining three. Never a zero.

AUTHORITY

—

no Search Console data

AUTHORITY0

No Search Console data — authority is link-structure only until you connect it.

HOVER AN EM-DASH FOR ITS REASON

ACT IIMaking the thing

05Research Studio

The model writes the brief. Code deletes what your evidence can’t prove.

Five specialist lanes fan out on one topic in parallel — keywords, live results, AI answers, opportunities, evidence — and you keep or discard sources while the run streams. A sixth agent writes the brief only after a server-owned readiness gate clears. Then five deterministic validators re-read the finished document: any currency or percentage figure not backed by evidence a human pinned is cut out of the sentence before you ever see it.

One brief invented a budget range roughly a fifth of what the session’s own kept evidence carried — the evidence said the equivalent of $40K–$500K, the brief wrote $8K–$20K. That span can no longer survive generation.
Numbers can ground only against evidence you pinned — the model is barred from citing its own generated evidence as proof of its own claim, which is the self-certifying loophole that makes an ungrounded figure look sourced.
It excises the span, not the bullet: an unbacked figure becomes “a cost that varies by project scope”, so an evidence-poor topic still produces a readable brief instead of no brief.

A lane that didn’t complete is retried on its own for one credit instead of paying for the whole run again — keywords, live results, AI answers and evidence each re-run alone.

Validator pass — after generation

EXCISED — UNGROUNDED

Procurement for a mid-market rollout clears in a single quarter in most of the sessions we kept evidence for. Teams should budget $8K–$20Ka cost that varies by project scope for the pilot phase, with integration effort the main variable, and hold the remainder against scope discovered in the first audit.

FIGURESPERCENTSOURCESQUOTESCLAIMS

THE POOL NUMBERS CAN GROUND AGAINST

PINNED · vendor pricing sheet — $40K–$500KPINNED · analyst cost noteGENERATED · barred from self-grounding

06AI Content Editor

Below the bar it’s refined or held — never shipped on a shrug.

Every draft is scored on four axes at once — SEO, GEO, E-E-A-T and Brand Voice — against the pages already ranking, and a Quality Gate sits between the draft and the publish button. A draft under the bar you set for that site is held, not published, and it comes back naming the exact signals that failed. An editor can clear the gate deliberately with a confirmation — it’s a gate, not a lock. And when the model attributes a statistic to a source that appears nowhere in the harvested research, the attribution is stripped or flagged before the draft is ever saved.

The failing signals are bars, not vibes: an answer-first intro of sixty words or fewer, a single H1 with no skipped levels, one external citation — and, on any article briefed at eight hundred words or more, at least two attributed statistics and one attributed quotation.
The Humanizer will refuse its own rewrite — a paragraph that loses a keyword or gains a link it shouldn’t have reverts to your original, and if the whole pass costs more than five SEO points every accepted paragraph is rolled back.
Autopilot drafts from four topic sources on a cadence you choose, but auto-publishing is off by default, needs an Owner or Account Admin to consent, and every grant and revoke lands in an append-only audit trail that is re-read at publish time.

Quality gate — between draft and publish

DRAFT · “Choosing a fleet-tracking platform”

SEO 71GEO 64E-E-A-T 58VOICE 82

THE GATEYOUR BAR · 70 · override: editor confirmation

PUBLISHED

no draft crossed
below the bar

HELD — QUALITY

intro answers at word 84 · bar ≤ 60

heading levels skip H2 → H4

external citations 0 · bar ≥ 1

attributed stats 1 · bar ≥ 2 at 800+ words

07Find it, fix it, take it back

A finding you can’t act on is just a complaint.

Every issue this platform surfaces is routed to a fix, not filed as a complaint — each one arriving with the move that fixes it, written by AI, previewed as a before-and-after diff, applied to your CMS in one click, and reversible. Then the next crawl re-checks the page: an issue counts as resolved only when the crawler can no longer find it on that URL — not when you ticked a box.

One write hits every major WordPress SEO plugin at once — Yoast, Rank Math, All in One SEO and SEOPress keys are set in the same request — and up to a hundred staged fixes go out in a single bulk action.
Where it can’t write, it says so: platforms it can only advise on get a deep link into their own editor and no Fix button at all, so no button ever produces a job nothing consumes. Shopify changes are always described as theme-dependent.
Rollback never claims success before it happens. A fix stays “applied” until the CMS confirms your original content was written back, and where we genuinely can’t undo something we name the door you have to walk through instead.

Approval is the default on every path — a fix, a draft, an outreach email all wait for a person. The one exception is Content Autopilot’s auto-publish, which stays off until an Owner or Account Admin turns it on for that site — and even then a draft under the bar is held, not shipped.

One fix — staged to reversible

STAGED→PREVIEW→APPLY→↺ UNDO

BEFORE

<title> Home — Acme Industrial

AFTER

<title> Torque Calibration Tools for Site Crews — Acme Industrial

<title> Home — Acme Industrial

YOASTRANK MATHALL IN ONE SEOSEOPRESS

ONE WRITE · EVERY KEY SET IN THE SAME REQUEST · THE LOOP ALWAYS RETURNS TO THE BEGINNING

ACT IIIThe work queue

08Opportunities

A scored board of your next moves — and a reason for every empty lane.

…so the first question of the week is already answered: what to work on. Your own site’s signals become a board of four lanes, each card carrying a fit score, the reason it surfaced and the specific next action. It populates from your own domain before you connect anything — ranking keywords, decaying pages and AI-citation gaps are the free first-scan set, so Search Console is an extra source rather than the price of entry. And a source that failed says so on the board, with its own state and its own count.

Every discovery source sits on a health bar with a live count and a verdict: 126 active failed 3 days 0 — index still thin for your domain not run yet tech sites only
The card’s button changes with the opportunity — Refresh Article, Optimize Page, Refresh Meta, Build Location Page, Plan Cluster — so the decision left to you is whether, not what.
Publish what it suggested and the outcome receipt has four distinct verdicts: measured, no match, no Search Console, or measurement failed. A missing rank is never rendered as “+0 clicks”.

An engine answering a question in your category and citing someone else becomes a card on this board like anything else — never cited, or cited before and not any more — scored by the same fit model as a decaying blog post. A prompt where nobody is cited isn’t counted as a gap; that’s whitespace, not a competitor beating you.

Discovery sources — live states

SOURCE SPINELAST SWEEP 09:12:44 UTC

Ranking keywords09:12126 active
Decaying pages09:1238 active
AI-citation gaps08:4712 active
Search Console06:03156 activefailed 3 daysFix →
Keyword index—0 — index still thin for your domainnot run yet
Competitor snapshots05:4024 active
Digital PR index05:12tech sites only
Audit findings04:5841 active

AN EMPTY LANE WITH A REASON BEATS AN EMPTY LANE.

09Keyword Research

Difficulty scored against your domain — with the evidence attached.

You picked a topic. The next question is whether you can actually win it. Every keyword tool hands the whole market the same difficulty number and asks you to trust it. This one scores each keyword against your site, publishes the arithmetic it used, and opens a drawer showing the exact results, the CTR gap and the citation count behind the verdict. When it can’t score honestly it prints “Not yet scored” — never a zero, never a blank cell.

Five weakness heuristics with published budgets — 30 · 30 · 15 · 15 · 10 — summing to exactly 100, and a heuristic that can’t be evaluated is renormalised out rather than quietly capping your score at a ceiling you can’t see.
We moved the winnable bar from 60 down to 38 because we measured 166 real scored keywords across four sites and nothing could reach 60. The maximum was 56.
A result page with fewer than four organic listings is refused a score outright — two forums in a two-result page is insufficient evidence, not a hundred.

Difficulty — arithmetic shown

THE WINNABLE BAR · 38

LONG SHOTWINNABLEWIDE OPEN

FORUM & UGC PRESENCE30 → 35

DOMAIN-VS-DOMAIN AUTHORITY30 → 35

CTR GAP15 → 18

STALE RESULTS15

no publish dates in this result page — renormalised out, the four remaining budgets refill the missing 15

CITATION COUNT10 → 12

10Rank Tracker

Your rankings moved. The volatility index tells you whose fault it is.

You shipped it and something moved. The next question is whether it was you. Checks run on an overnight schedule at 02:00 UTC — how often yours run depends on your plan. What doesn’t vary is what happens when a check comes back empty: a keyword that drops out of the results still gets a history row, so rank loss shows up as a drop rather than a gap in the chart. And a published 0–10 volatility index, calibrated so a typical night lands on 3, separates a market-wide algorithm storm from a drop you caused.

Unranked is never zero. A missing position is excluded from every average, and a mover’s delta is null rather than invented — the same sentinel enforced across four separate readers after it bit us once.
Below its sample threshold a reading is withheld and labelled “collecting”, not “low” — and the comparison tile returns “Still collecting enough data to compare your site against the market” instead of defaulting to “in line”.
Below fourteen days of history the index refuses to calibrate off a noisy short window and falls back to a published reference point instead, and a whole category row is withheld until it carries five hundred movement samples.

Volatility index — thirty nights

IN LINE WITH GLOBAL — ALGORITHM STORMABOVE GLOBAL — LOOK AT YOUR SIDE

GLOBALYOUR SITETOLERANCE ±1.0

11Competitive Analysis

Point at the page that’s losing. Get a scored gap and the brief.

It wasn’t the market, it was one page. Point at it. Everyone else answers “how big are they?” — domain rating, estimated traffic, backlink counts — and leaves you to guess what to edit. This answers “why is their page beating mine, and what do I change?” You point a deep dive at one of your own URLs, it reads your page alongside the pages actually ranking, and it returns a topic-by-topic coverage matrix, a scored gap with every finding attributed to the competitor it came from, and a one-click brief. Then plain deterministic code re-reads the report and deletes what it can’t support.

A “missing” topic that matches a heading already on your page is dropped, never repaired — and so is a “gap” whose rationale praises your page instead of naming a deficit.
The coverage matrix has a fifth cell state. A competitor page we couldn’t fetch reads “Couldn’t read” and gets named in an exclusion note — never counted as Missing, which is the free win that would have flattered your score.
Page Quality can’t read 100 while the report below it lists problems: every flagged thin section or missing topic docks the score, capped at a 30% reduction, so the number and the findings can’t contradict each other.

Coverage matrix — deep dive

Topic

YOUR PAGE

fieldkit.io

loopware.co

brightpath.dev

stackrow.net

Pricing breakdown

Strong

Moderate

Weak

Integrations

Strong

Moderate

Strong

Moderate

Migration steps

Weak

Strong

Weak

Strong

Security & compliance

Moderate

Strong

Weak

Comparison table

Missing

Strong

Moderate

Implementation timeline

Moderate

Weak

Moderate

Strong

FAQ schema

Strong

Missing

Weak

Moderate

Case studies

Weak

Strong

Moderate

Strong

TCO analysis

Missing

Moderate

Strong

Weak

COULDN’T READ

excluded from the comparison and named — its coverage is unknown, never counted as Missing

12Backlinks

Outreach from your own inbox — and a pitch that can’t invent a number.

The page is fixed and still losing. That’s usually links. Link gaps come from cross-referencing your rivals’ profiles against yours, and the outreach runs from your own Gmail or Outlook on a send-only permission — sending never asks for read access to your mailbox. Before a cold email sends, a deterministic guard deletes every sentence containing a number or an engine name the evidence doesn’t back. “We have helped 240 brands grow their organic traffic” is removed from the draft, not flagged. And nothing AI-drafted leaves your mailbox until a human approves it — the server refuses an unapproved send outright.

Link gap is arithmetic you can check — each prospect’s likelihood is the share of your tracked competitors linking to it × 80, +10 if its authority is high, +10 if it links to more than one of them, clamped 0–100. You get a prospect list, not a blank search box.
It tells you which page needs links, not that you need links — high-value pages with no inbound links are named individually. Referring domains lead on link impact — the visitors a domain actually sent, weighted by its authority — so the high-authority social profile that sent six people stops outranking the niche blog that sent four hundred.
If the guard strips so much that the draft no longer cites real evidence, the pitch is discarded and the credit refunded — rather than quietly sending a generic email that calls itself personalised.

We don’t promise you a set number of links a month. No link exchanges, no automated link creation, no guaranteed counts — all three are named in Google’s own spam policies. And the At Risk view says in its own subheading what it is not: link-health risk, not a toxic or spam score.

Outreach guard — the evidence ledger

“Your resources page links to a calibration guide that moved last spring.”
DOCKED — BACKED

“Our torque calibration guide covers the step your readers currently lose.”
DOCKED — BACKED

“We have helped 240 brands grow their organic traffic.”
REJECTED AT THE EDGE

EVIDENCE — WHAT THE PITCH MAY CITE

broken link on their resources page → 404your matching asset: /torque-calibration-guidethey link to two of your tracked rivalsno traffic claims pinned1 UNBACKED CLAIM DROPPED

SENTENCES DOCK AGAINST THE LEDGER OR THEY DON’T SEND. NOTHING STRIKES THROUGH.

13Brand Integrity Monitoring

AI is describing your brand. Check every sentence against facts you signed off.

And while you were doing all of that, AI was describing your company to buyers who never reached the site. Classic brand monitoring answers who said your name and in what tone — it has nothing to say about whether the claim was true. Here you keep an editable fact sheet of what AI should get right about you, including one category that runs in reverse: the things that are not true of you, like a certification you don’t hold or a market you don’t serve. When scanning is enabled for a site, every statement an engine makes about your brand is graded against that sheet — accurate, outdated, wrong, or unverifiable. Available on Pro and above.

If the quoted sentence isn’t an exact substring of the answer the engine actually gave, the finding is thrown away — not softened, not repaired into a match. Every candidate the parser can’t match verbatim is discarded before it can be scored.
The grader is told to reserve “critical” for five triggers — a direct contradiction of founding year, price, leadership or location, or a forbidden-claim violation — so red is scoped rather than sprayed.
A brand that has never been scanned is never displayed as accurate — checked and clear, not enabled yet and no scan history are three separate messages the product ships.

Brand facts — the case file

FACT SHEET · F-07

Founded 2016

source: /aboutCONFIRMED

REVERSE FACT · not true of us: certifications we don’t hold

EXHIBIT · CHATGPT ANSWER · SAMPLED 08:41

“…founded in 2011, the company builds an SEO and AI-visibility platform for small teams…”

WHAT THE AI SAID“founded in 2011”CRITICAL — FOUNDING YEAR

YOUR FACT SHEET SAYSfounded 2016 · confirmed · /aboutCONFIRMED

SUGGESTED ACTIONpublish the correction snippet on /about · re-checked on the next scan1-CLICK

accurateoutdatedwrongunverifiable

14llms.txt Generator

We measured whether it works. It didn’t move the needle.

Every other tool sells llms.txt as an AI-visibility lever. We ran the before-and-after on our own monitored accounts and found that a change in answer-engine citation rate within ninety days of publishing was not distinguishable from noise. So we put that on the product page instead of the pitch, took it off our own list of AI-citation tactics, and cut the price to zero — the metered bucket was deleted from billing, not zero-rated. What’s left is an honest publishing utility that costs about two minutes.

Free generators hand you a skeleton with the summaries left blank. That blank is the entire job, and it’s the part we write — in your trained brand voice, capped at twenty-five words, strictly factual, with your banned phrases enforced.
“Published” isn’t “live”. The status only reaches Live & verified after we fetch the file back from your own domain, and a 200 that returns an HTML shell is rejected rather than counted — the false positive every other checker falls for.
Four delivery paths, two automatic and two we label as guided, because they are not one-click and we don’t say they are.

Generate one right now — no account, no card, just an email address to send the file to.

llms.txt — the measurement

~2 MIN SETUP0 CREDITS0 MAINTENANCE

“a clean, structured read for the engines that use it”

NO DETECTABLE CITATION EFFECT

REMOVED FROM OUR AI-CITATION TACTICS LISTDE-METERED · 0 CREDITSMOVED OUT OF THE CONTENT SECTION IN-APP

15Agencies and multi-site teams

Every client keeps their own everything.

A workspace holds brands, a brand holds sites, and each site carries its own connections, its own brand voice, its own fact sheet, its own opportunity board, its own keyword set and its own audit history. Nothing is re-entered when you switch between them. Every tenant-scoped read carries an explicit workspace predicate, and switching clients switches the whole context — not just a filter on one screen.

Four ordered roles — Owner, Admin, Editor and Viewer — with Editors scoped to write only on the sites you assign them and Viewers read-only across the workspace.
Integrations are per site, not per account: Search Console, Analytics, Ads, your store and your CMS are each connected on the site they belong to.
Auto-publishing consent is a role gate, not a checkbox — only an Owner or Account Admin can grant it, an Editor is refused outright, and every grant and revoke is recorded in an append-only trail that gets re-read at publish time.

Workspace — tenant containment

OWNERADMINEDITORVIEWEReditor: write on assigned sites only

WORKSPACE · YOUR AGENCY

bramble-and-birch.com

SCGACMS

CMS write · read-only for Editor

fieldnotes-supply.com

SCGASTORECMS

assigned · Editor writes here

stonewharf-hotels.com

SCCMS

CMS write · read-only for Editor

QUERY · WORKSPACE-SCOPED

THE PULSE RETURNS TO ITS OWN CARD — THE NEIGHBOURS STAY DARK.

16One subscription, one loop

The work hands itself off — and names its price before you click.

Research produces a brief. The brief produces a draft. The draft meets the gate. The approved fix goes to your CMS. The next crawl checks whether it landed, and the outcome receipt tells you which of four things it could actually measure. Nothing in that chain spends without telling you first, and every publish waits for a person unless an Owner or Account Admin has explicitly switched auto-publish on for that site.

The price is on the button before you press it — “Free to claim”, “Run deep dive · 1 credit”, “Analyze this keyword · 1 credit” — and manual entry is free and says so.
Budgets are per feature rather than one opaque pool, so running out of drafts doesn’t stop your audits, and a refusal names the feature that ran out instead of showing a bare “out of credits”.
A failed run is refunded by the same code that charged it, so a charge and its reversal can’t disagree about the amount.

A weekly summary lands in your inbox with what moved — and it omits a line rather than reporting a zero. A quiet week produces a shorter email, not a reassuring one.

Credits — the receipt

Run deep dive1 CREDIT−1

SEARCHCHAMP · CREDIT RECEIPT

Deep dive · rival page vs yours−1

Content brief · from the gap report−1

Keyword analysis · run failed−1

Keyword analysis · refunded+1

BALANCE THIS SESSION−2

REFUNDED — RUN FAILED

DEEP-DIVE BUDGET

PER-FEATURE BUDGETS · PRICE SHOWN FIRST · REFUND BY THE SAME CODE THAT CHARGED IT

Plans

Compare capacity across every plan

Every capability is on every plan — capacity is the difference: more sites, seats, keywords, and faster refresh as you move up. Every plan is paid; there is no free tier.

Feature	Starter	Pro	Agency
Sites	2	5	15
Team seats	1	3	10
Tracked keywords	100	1,000	3,000
AI-visibility prompts / mo	30	100	250
AI-visibility engines	5	5	5
AI content articles / mo	10	40	100
Audit pages / mo	2,000	25,000	100,000
Rank tracking refresh	Weekly	Daily	Daily
AI-visibility refresh	your top prompts daily, the rest weekly	your top prompts daily, the rest weekly	your top prompts daily, the rest weekly
AI agent chat	Unlimited	Unlimited	Unlimited
Backlinks + outreach	Built-in outreach CRM	Built-in outreach CRM	Built-in outreach CRM
Daily rank tracking	—	✓	✓
API + full MCP toolset	—	✓	✓
White-label reports	—	—	✓
Multi-market prompt tracking	—	✓	✓

Frequently asked

Common questions about SearchChamp

Hover or click a question for the answer

The core toolset, yes. Keyword research, rank tracking, site audit, AI content, AI visibility, backlinks, opportunities, and llms.txt are bundled on every paid plan from Starter (59) up, and you pick a plan mainly by scale — sites, seats, monthly allowances. Publishing to WordPress, Shopify and the other connected platforms, and autopilot publishing, are on every paid plan too. A defined set of advanced capabilities does sit higher up, and we'd rather list them than bury them: API access with the full MCP toolset, AI shopping visibility, personas, multi-market tracking, local-grid tracking and brand-integrity scanning are on Pro and up; white-label client reports are included on Agency; SSO/SAML with an SLA is Enterprise.

GENERAL

Almost never — and never for the core toolset. Keyword research, rank tracking, site audit, AI content, AI visibility and all seven agents are bundled on every paid plan, and top-ups only add volume when you pass a plan's monthly allowance. Capabilities are tied to plans rather than sold as separate add-ons: white-label client reports are part of the Agency plan, and a set of advanced capabilities sits on Pro and up — API and the full MCP toolset, AI shopping visibility, personas, multi-market and local-grid tracking and brand-integrity scanning — with SSO/SAML and an SLA on Enterprise. Publishing (WordPress, Shopify and the other connected platforms) and autopilot publishing are on every paid plan.

GENERAL

Six specialist AI agents — Keyword Research, Content Writer, Site Auditor, AI Visibility, Brand Voice, and Opportunities — coordinated by a Concierge orchestrator. They run the full crawl-to-cite loop: find what to target, audit what's broken, draft and publish content, then track citations across the AI answer engines (ChatGPT, Google AI Overviews, Gemini, Google AI Mode, and Microsoft Copilot) on every plan and surface what to fix next. A free llms.txt Generator is also included on every plan.

GENERAL

Live now: AI Visibility across AI answer engines, Keyword Research, Site Audit (80+ checks), AI Content Writer, Rank Tracker, Backlinks (monitoring, competitor gaps, outreach, and recovery), Opportunities, llms.txt, Brand Voice, Competitive Analysis, Research Studio, the WordPress plugin, the public REST API (Pro and up), and GSC / GA4 / Google Ads / Shopify integrations. Coming soon: the Chrome extension. We label roadmap items clearly — nothing in Beta is presented as shipped.

GENERAL

5 AI answer engines on every plan — ChatGPT, Google AI Overviews, Gemini, Google AI Mode, and Microsoft Copilot. For GEO scoring in the Site Auditor, SearchChamp also checks how five AI crawlers — Googlebot, GPTBot, ClaudeBot, PerplexityBot, and GeminiBot — read your pages.

GENERAL

Yes. AI-visibility citation tracking is included on every paid plan: all plans cover the same 5 AI answer engines (ChatGPT, Google AI Overviews, Gemini, Google AI Mode, and Microsoft Copilot). Starter includes 30 tracked prompts per month; Pro 100; Agency 250.

GENERAL

Yes — the data exports are compatible. Most users who try SearchChamp as a complement find they replace their main tool within 60 days, primarily because AI visibility tracking and the AI content writer are included instead of charged as extras.

GENERAL

Yes — a 7-day free trial on the Starter plan, credit card required. At the end of the trial, your card is charged for the plan you selected at signup. Cancel anytime during the trial and you won't be charged.

GENERAL

Yes — a credit or debit card is required at signup. You won't be charged until the 7-day trial ends, and you can cancel anytime before then with no charge.

GENERAL

Our SEO data engine for keyword + SERP + audit data (same source as several incumbent platforms), Google Search Console for first-party site data, and PageSpeed Insights for performance.

GENERAL

Yes — citation history exports for reporting, so an agency doesn't have to screenshot the dashboard to prove a client's AI-answer visibility moved. It carries the same detail as the in-app view: which engine, which query, cited or not, and the position when cited.

CAPABILITIES

On a scheduled cadence tied to your plan rather than continuously — each check queries every engine for every tracked prompt, and running that too aggressively would mistake day-to-day answer variance for a real signal shift. Your dashboard shows exactly when the last check ran and when the next one fires.

CAPABILITIES

Brand Voice trains per site on the content you provide it, so a site publishing in more than one language trains a voice profile per language rather than translating a single English voice — the goal is content that reads as genuinely native, not machine-translated.

CAPABILITIES

It regenerates automatically on publish, so a site that updates often stays current without anyone remembering to re-run it manually. For content that changes outside a normal publish event (pricing tables, live inventory), the summary reflects the state at the last regeneration — treat it as close-to-current, not real-time.

CAPABILITIES

The Chrome extension is free for any paid SearchChamp account — no separate install purchase. It gives you keyword volume, difficulty, and AI Overview presence data for any page or highlighted text, without leaving your browser.

INTEGRATION & SCALE

Yes — it's included with SearchChamp, not a future roadmap item. It auto-applies SearchChamp audit recommendations (meta tags, schema, and more) directly from the WP dashboard, publishes content, and scores posts live inside Gutenberg.

INTEGRATION & SCALE

One invoice covers every site in the workspace, with the monthly allowance shared and flexible across them rather than metered per site — so budget moves to wherever the work actually is that month instead of sitting reserved on a quiet client.

INTEGRATION & SCALE

Every core capability — keyword research, rank tracking, site audit, content editor, AI Visibility, backlinks, and the Opportunities board — is on every paid plan starting at Starter. What changes by tier is volume: how many sites, how many tracked keywords, how much content allowance, and (on Agency) multi-client workspace management. Nothing on this page is a Pro-or-above exclusive unless stated.

PRICING

Pro and above. It's how teams build internal dashboards or automations on top of their SEO data without a custom integration project — included in the plan rather than a separate enterprise line item.

PRICING

Every score on this page publishes its arithmetic

Start free trial See pricing

7-day free trial · credit card required · cancel anytime