PondralFrequently asked questions.
Straight answers about what AI Visibility is, how Pondral measures it, which engines and models it checks, what each plan costs and includes, how your data is handled, and how accurate the scores are. Every number here matches the pricing page and the methodology page; if you find one that does not, email hello@pondral.com.
What AI Visibility is
What is AI Visibility?
AI Visibility is whether AI assistants such as ChatGPT, Claude, Gemini, Perplexity, and Grok name, describe, and link to your brand when buyers ask them for options. It is measured per question and per engine, because the same brand can be recommended by one assistant and ignored by another on the same day.
How is AI Visibility different from SEO?
SEO measures where your pages rank in a list of links. AI Visibility measures what an assistant says when it gives one synthesized answer instead of a list. You can rank first on Google and be absent from the ChatGPT answer to the same question, and the reverse. The two share foundations (crawlable pages, clear entity facts, third-party mentions) but are scored differently. See pondral.com/ai-visibility-vs-seo.
Is AI Visibility the same as Answer Engine Optimization (AEO)?
No. AI Visibility is the outcome you measure. Answer Engine Optimization is the work agencies and content teams do to improve it. Pondral measures the outcome and shows which questions to work on.
Which brands need to care about this?
Any brand whose buyers ask an assistant for recommendations before they search: B2B software, agencies, professional services, consumer products with a research step. If your category question returns a list of names and yours is not in it, that is lost demand you cannot see in analytics.
What does Pondral actually do?
Pondral asks the 5 AI engines a fixed set of your buyers' questions every week, stores every answer, scores each on a published five-factor rubric, and shows you which questions you are missing from, which competitors are named instead, and what to fix.
Who is Pondral for?
In-house marketing teams tracking one or a few brands, and agencies tracking many client brands. Plans start at $149 a month for one brand.
How Pondral measures it
What is the AI Visibility Score?
A 0 to 100 number that is the average of every graded answer in an audit run. Each answer is graded on five factors: Presence (20%), Prominence (25%), Context (20%), Citation Link (20%), and Competitive Presence (15%). The full method is at pondral.com/methodology.
What do the five factors mean?
Presence: whether the brand appears in the response at all. Prominence: how early in the response the brand appears. Context: how well the brand is described, portrayed favourably, described accurately, and placed in the right category. Citation Link: whether the response cites a link to your own domain. Competitive Presence: your share of the brand mentions in the same response.
Which questions does Pondral ask?
The buyer questions you choose for each brand, built in the Question Panel builder or added one at a time, drawn from four templates: category entry, problem first, comparison, and brand adjacent. Free and SMB brands carry up to 10 questions, Growth 50, Agency 100.
How often does Pondral run an audit?
Every paid brand gets a full 5-engine audit weekly, and you can run one on demand from Monitoring. Growth and Agency plans also get daily change alerts (2 engines, first 2 questions), so a sudden disappearance is caught between weekly runs.
Why does my score change between runs when I changed nothing?
AI engines are non-deterministic. The same question can return a different answer an hour later. We measured this: running the same audit twice on the same day moved the overall score by 0.59 points and individual answers by 1.86 points on average (PROXY, July 31, 2026). A few points of movement is noise; 15 or more on a named question is signal.
Can I see the actual answer behind a score?
Yes. Every score links to the stored answer text, the engine, the model version, the timestamp, and the cited URLs. The Reports page exports the full evidence for any run as CSV or JSON.
What is a gap?
A buyer question where your brand is absent from, or ranked low in, an engine's answer. Gaps are ranked by how much they would move your score if closed, and each one can become a writing brief.
Does Pondral tell me what to fix?
Yes. Each gap carries a suggested fix and can generate a brief for the page that would answer that question. Whether the fix worked shows up in the next weekly run.
Does a higher score mean more customers?
Not by itself. The score measures what assistants say, not what buyers do next. Pondral is measurement; the commercial result depends on what you do with it. We make no customer-outcome claims.
Engines and models
Which AI engines does Pondral check?
ChatGPT, Claude, Gemini, Perplexity, and Grok. All five are included on every paid plan; there are no engine add-ons and no enterprise-only engines.
Which model versions are used?
As of September 22, 2026 (methodology 2.7.0): GPT-5.5 for ChatGPT, Claude Sonnet 4.6, Gemini 2.5 Flash, Perplexity sonar through the Agent API, and Grok (Grok 3 on self-serve monitoring audits, Grok 4.3 on consulting audits). When a vendor changes a model we record it as a methodology version and never average readings across the change.
Does the free check run all five engines?
No. The free check runs 2 engines per check, 3 checks a month, on a fixed 12-question mini panel, in about a minute and with no signup. It is a directional snapshot, not a monitoring instrument.
Why do the engines disagree about my brand?
Because they draw on different sources. ChatGPT and Claude lean on training memory plus web search; Gemini follows Google's index; Perplexity retrieves a small set of live pages and weights community and review sites heavily; Grok reads X first. A brand can be named by ChatGPT and absent from Perplexity on the same question.
Does Pondral force the engines to search the web?
Yes, wherever the vendor allows it. An answer that skipped web search is discarded and re-run, so your score reflects what a grounded assistant says today, not what a model remembers from training.
Will you add more engines?
An engine is added when it has meaningful buyer usage and an interface that can be grounded and stored. Additions are recorded in the methodology changelog before they affect any score.
Why did my Perplexity score change on September 21, 2026?
Perplexity retired the interface we used and we moved to its Agent API and sonar model (methodology 2.7.0). Readings before and after that date are stamped with different versions and are not averaged together.
Pricing and trial
How much does Pondral cost?
Free: $0, 3 free checks a month on 2 engines. SMB: $149 a month, 1 brand. Growth: $449 a month, 6 brands. Agency: $999 a month, 12 brands. Annual billing is 15% less. Extra brands on paid plans are $75 a month each.
What counts as a brand?
One company or product you monitor, with its own website, competitors, and question panel. An agency tracking twelve clients needs twelve brands.
Is there a free trial?
Yes. Every paid plan starts with a 14-day free trial. A card is required at signup, and the plan price is billed automatically when the trial ends. Cancel any time during the trial from your billing settings and you will not be charged. The Free tier is permanent and needs no card.
Can I cancel any time?
Yes, from your billing settings, effective at the end of the billing period. Cancelling during a trial means you are never charged.
What happens when I hit my plan's limits?
We never auto-charge overages. Each brand tracks up to its plan's question allowance (10 on Free and SMB, 50 on Growth, 100 on Agency), and monthly checks stop when the allowance is used up. Upgrade whenever you want more; nothing is billed automatically.
Can I add more brands to a plan?
Yes, on any paid plan, at $75 a month per extra brand, up to 3 total on SMB, 10 on Growth, and 25 on Agency.
Do you offer white-label reports?
Yes, on the Agency plan: branded reports, a shared client portal, and an evidence pack you can put in front of a client.
Is there an API?
Yes, on Growth and Agency. Documentation is at pondral.com/docs/api and keys are created in Settings.
Privacy and data
Where is my data stored?
In the US (us-east-1) on Supabase, with hosting on Vercel. The full list of providers that process data is at pondral.com/sub-processors.
Do you sign a data processing agreement?
Yes. The DPA at pondral.com/dpa includes the EU standard contractual clauses and names every sub-processor. To put it in place for your account, email privacy@pondral.com with the subject "DPA Inquiry". Pondral does not offer healthcare or financial-institution plans and no Business Associate Agreement is available.
Do you train AI models on my data?
No. Your questions, answers, and scores are used to produce your reports and for nothing else. Details at pondral.com/legal/privacy/training-data.
Which cookies do you set?
Essential cookies for sign-in, and optional analytics tools (PostHog and Sentry) that load only after you accept the banner. There are no advertising trackers.
Can I delete my account and data?
Yes, from Settings, under Danger zone. Deletion removes your account and Pondral-held data and cancels any active subscription; billing records are kept as the law requires. Question text already sent to the AI engines at audit time is governed by those vendors' retention policies and cannot be recalled by Pondral. Audit data on a live account is kept for 90 days on Free, 180 on SMB, and 365 on Growth and Agency, then archived for 30 days and deleted.
Are you SOC 2 certified?
Not yet. We have not completed a SOC 2 audit or an external penetration test. Our security practices are described plainly at pondral.com/security.
How fast do you notify customers of a breach?
Within 72 hours of confirming it, as written in the DPA.
Accuracy and honesty
How accurate are the scores?
Four of the five factors are rule-based and exact for a given answer. The fifth, Context, is graded by a model. Measured in August 2026 on 150 randomly drawn real engine answers, that grader agrees with a different-provider grader on 87.3% of cases (Cohen's kappa 0.624) and with a larger same-provider grader on 72.0% (kappa 0.402). We publish both because they disagree. These are single measurements and are directional (PROXY). Whole-run stability is a separate measurement; see the question about scores changing between runs.
Have you ever published a number you had to withdraw?
Yes. On July 27, 2026 we withdrew a published grader-agreement figure (kappa 0.947) because it had been measured on sentences we wrote ourselves rather than on engine answers, and a calibration figure whose script counted failed grader calls as matches. The record and the replacement figures are at pondral.com/blog/methodology-integrity. Every change to the method since is in the methodology changelog.
What do PROVEN, PROXY, and ILLUSTRATIVE mean on your pages?
PROVEN: measured on production data with the method described. PROXY: measured, but on a limited sample or a single day, so it indicates rather than proves. ILLUSTRATIVE: a worked example, not a measurement. The label stays attached to the number wherever it is quoted.
Do you have customer case studies?
Not yet. Every number we publish is Pondral's own measurement: our audits of ourselves and the public teardowns at pondral.com/blog/audits.
Can I reproduce a score myself?
Yes. The rubric, weights, question templates, and model versions are public, and every audit ships its raw answers with timestamps. Ask the same engine the same question with web search on and you should get an answer that grades within the published variance.