Best AI Visibility Tools Compared: Profound, Peec, Otterly and DIY

An AI visibility tool sends a fixed set of prompts to one or more AI engines on a schedule. It records which brands the answers name and which pages they cite, and charts the result over time. Profound, Peec AI and Otterly sell that as a subscription.

A developer with an afternoon can build the same loop against the OpenAI, Perplexity and Gemini APIs plus a SERP provider for a few dollars a month in fees. This guide compares the three products and the DIY route on what their own pages state. It prices the API route at the sample size the AI Shortlist Index uses, and gives a checklist for evaluating any of them.

Every vendor fact here comes from a page fetched on 2026-09-04; anything a vendor does not publish is marked as such.

What does an AI visibility tool measure?

Under the dashboards, every tool in this category holds a prompt list and runs each prompt against each engine it supports on some cadence. It then extracts brand names and cited URLs from the answers and stores the runs so a chart can be drawn.

The prompt list

The prompt list is the part that decides whether the numbers mean anything. All three products let you write your own prompts, and each also sells prompt discovery.

Profound’s “Prompt Volumes” claims to show what people ask AI, and Peec’s onboarding is built around setting up the prompts that matter. Otterly has a “Query Fan out Tool” that lists the sub-queries an engine might generate from one prompt.

Discovery is useful for finding phrasings you had not thought of, but it does not tell you which phrasings a property manager with 300 doors uses. A prompt set that skews toward generic “best software” questions will overstate whoever has the most marketplace listings. The measurement framework covers how the Index builds a set across six prompt types so one phrasing cannot decide the result.

Engines, citations and sentiment

Engine coverage is the second variable and the one the pricing pages sort tiers by. Profound’s homepage lists eight engines, but its $99 tier tracks ChatGPT only and its $399 tier tracks three.

Otterly includes four engines on every plan and sells Claude, Google AI Mode and Gemini as add-ons. Peec’s homepage shows ChatGPT, Perplexity and Gemini.

Citations are the third output, and the one worth paying for. All three state they record the sources an answer cites: Profound as “Source Citations”, Peec as “Find Key Sources”, Otterly as domain and URL citations checked daily. A citation list tells you which pages carry your mentions, and those pages are where the work is.

Sentiment is the fourth. Otterly describes its version as sorting mentions into recommended enthusiastically, mentioned with caveats, or dismissed outright. Read any sentiment score as a coarse framing signal, not a review score.

What does no tracker measure?

None of these tools sees a click. An API answer or a scraped AI Overview sends no traffic anywhere. None of the OpenAI, Perplexity, Gemini or SerpApi documentation cited in this guide describes a way to see what a real user did after reading an answer.

A tracker can tell you that you were named on forty prompts out of a hundred and cited from a G2 page. It cannot tell you whether anyone booked a demo. The gap between a mention count and a lead is wider than the gap between a rank and a lead, because the buyer may never leave the answer.

They also do not see the model version behind each run unless they choose to store it. The Index protocol stores the model identifier with every run where the engine discloses one, because the engines change models without notice. A chart that moves on the day a model changed is not a chart of your work.

None of the three states on the pages fetched for this guide that the model identifier is stored per run; ask.

How do Profound, Peec and Otterly compare?

Tool	Engines	Prompts per plan	Citations	Sentiment	Competitor view	Geography	Run frequency	Published price
Profound	ChatGPT only on Starter; ChatGPT, Perplexity, Google AI Overviews on the $399 tier; up to 9 on Enterprise (homepage shows ChatGPT, Perplexity, Claude, Gemini, Grok, Copilot, DeepSeek, Google AI Overviews)	50 / 100 / tailored	Yes, “Source Citations”	Yes	Not stated as a feature; Enterprise lists “multiple companies tracked”	1 region, 1 language on the two self-serve tiers	Not published	$99/mo and $399/mo, billed yearly; Enterprise custom
Peec AI	ChatGPT, Perplexity, Gemini	Not published	Yes, “Find Key Sources”	Yes	Yes, “Add Brands”	“All countries”	Not published	Not published; four tiers, annual billing, Enterprise by sales call
Otterly	ChatGPT, Google AI Overviews, Perplexity, Copilot on all plans; Claude, Google AI Mode, Gemini as add-ons	15 / 100 / 400 / custom	Yes, domain and URL citations, daily	Yes	Yes, “Me + Top 5 competitors”	50+ countries (pricing page), 65+ countries and languages (features page)	Daily on Standard and Premium	$29 / $189 / $489 per month; $25 / $160 / $422 annual; Enterprise from $1,000/mo
DIY (OpenAI, Perplexity, Gemini APIs + SerpApi)	ChatGPT via the web search tool, Perplexity Sonar, Gemini with Google Search grounding, AI Overviews via SerpApi	Whatever you write	Yes, if you store them	Only if you build it	Only if you build it	Whatever the API parameters allow	Whatever you schedule	Metered; under $5/month in API fees at 25 prompts by 3 runs by 4 engines, before engineering time

From each vendor’s own pages fetched 2026-09-04: Otterly publishes the most detail, including daily runs and a $29 entry price. Profound publishes two self-serve tiers at $99 and $399 with one region each. Peec publishes tier names but no prices, prompt counts or cadence, and the DIY route costs the least in fees and the most in hours.

Three things stand out. Only Otterly states how often it runs. Only Profound states a region limit, and its self-serve tiers have one.

Peec, whose homepage claims 3,000 brands and agencies, publishes fewer numbers than either competitor, so a buyer has to take the sales call to learn what a prompt costs. The pricing pages for Profound, Peec AI and Otterly are the sources; check them before quoting a figure, since these pages change.

What does the DIY route cost?

The Index protocol is 25 prompts per category, three times each, per engine, over at least two days. That is 75 requests per engine per month per category. Priced against the documentation fetched for this guide, the API bill is small.

The arithmetic

OpenAI’s pricing page lists the web search tool at $10 per 1,000 calls with search content tokens billed at model rates, so 75 calls is $0.75 before tokens. Perplexity’s pricing page gives sonar at $1 per million tokens plus a per-request fee of $5 to $12 per 1,000 depending on search context size. So 75 requests is $0.38 to $0.90 plus cents in tokens.

Google’s Gemini API pricing page gives Gemini 3 models 5,000 free Google Search grounding requests per month and $14 per 1,000 after that. So 75 grounded requests cost nothing unless the quota is spent elsewhere.

SerpApi’s documentation says the AI Overview content comes from a second request using a page token that expires within one minute of the first search. Each AI Overview capture is therefore two searches, and 150 searches sits inside SerpApi’s free tier of 250 per month; the $25 Starter plan covers 1,000.

The real cost

Add it up and one category at the Index floor costs under $5 a month in fees, or under $30 with a paid SerpApi plan for headroom. A vendor tracking one category at three times the Index sample is still under $50.

The fees are the smallest line. The real cost is the code that extracts names with aliases and case rules, and the store that keeps every run with its model identifier and cited URLs. Add the review queue for names the matcher missed, and the report someone has to read.

What does DIY lack?

A DIY tracker lacks two things a commercial tool has, and only one of them matters.

SERP capture

ChatGPT, Perplexity and Gemini each expose a documented API with web search or grounding, priced on their pricing pages, and citations come back from those APIs as data. The Gemini API covers Gemini models and not Google AI Overviews. That is why the Index protocol captures Overviews through a third-party SERP provider and records “no AI Overview” when the query returns none.

A DIY tracker inherits that dependency plus the two-request page-token dance, and the SERP provider’s output format is the one thing in the stack you do not control. The guide to how AI Overviews are triggered and sourced covers why absence is the common result on B2B software queries.

UI parity

UI parity is the other, and it is not worth chasing. A commercial tool ships a dashboard, alerts, competitor overlays, a share-of-prompts chart and a place to invite the agency. A DIY tracker ships a table.

If the person reading the numbers is the person who built the pipeline, the table is enough. If the reader is a VP who wants a chart in a deck, the dashboard is what you are paying $189 or $399 a month for. That is a reasonable trade.

The consumer-product gap applies to both

Otterly’s homepage says outright that a manual search in the ChatGPT app may not match its results because of personalization such as memory and location. That is the honest position. An unpersonalized run measures the baseline; your own signed-in account measures one account.

Reading the app rather than an API does not close that gap. That is why the Index captures ChatGPT from the app with memory off and the chat set to Unpersonalized. Any tool that claims to measure “what your customers see” is measuring a baseline and calling it something else.

How do you evaluate any AI visibility tool?

Run these questions past a vendor, or past your own build, before paying for a year.

1. Does the prompt set represent real buyers? Ask to see the prompt discovery output for your category and count how many prompts a property manager would type. A set with problem-led prompts, integration prompts naming a specific PMS, and “alternatives to” prompts naming the incumbent measures the shortlist. The checker template has a prompt set shaped that way.
2. How many runs per prompt, and over how many days? One run per prompt per day produces a consistency value of nothing, because one run cannot disagree with itself. Three runs in fresh sessions across two days is the floor. None of the three vendors publishes a runs-per-prompt figure.

The engines and the geography

3. Which engines, and on which tier? Match the plan’s engine list to the engines your buyers use, not the homepage’s roster. Profound’s entry tier is ChatGPT only; Otterly’s Gemini is an add-on.
4. Which geography? Profound’s self-serve tiers state one region and one language. Otterly states 50 or more countries, and Peec states all countries. For US property-ops software the one-region limit is fine.

The data

5. Are citations stored per run, and exportable? A tool that shows the top cited domains this week but cannot hand you the URL list per prompt per run is showing a summary. It is data it will not let you keep, and Otterly lists API access at 2,000 requests a month on Standard; ask the others.
6. API model or consumer product? If the vendor scrapes consumer apps, ask which account, which location and whether memory is off. If the vendor calls APIs, ask which models and whether the identifier is stored per run.
7. Can you leave with your data? A year of runs is the asset. If cancelling means losing the history, the price is higher than the pricing page says.

Where does the AI Shortlist Index fit?

The Index is a published monthly measurement for property-ops software categories. Its protocol runs a frozen 25-prompt set three times per engine and stores every run with its cited URLs and the model identifier where disclosed. It publishes mention rate, position, consistency, engine spread and source dependency per vendor.

Today it measures ChatGPT and Claude; Perplexity, Gemini and Google AI Overviews have adapters and no runs. The methodology is public down to the extraction rules.

That makes it one option among the four in the table, and for a narrow buyer. If you sell into a property-ops category the Index covers, you get a monthly reading at a defensible sample without buying a tool or writing a pipeline.

What it is not

It is not the winner of this comparison. It covers property-ops categories only, runs monthly rather than daily, uses its own prompt set rather than yours, and will not add a prompt because you asked.

A vendor that wants its own prompts, cadence and competitors on a chart needs one of the other three routes. The alternative is the monitoring retainer, which delivers a per-edition movement report from the same run files, and paying for that never affects placement.

The first edition is a useful reminder of what a sample is. The 2026-09 export records 25 ChatGPT runs per category, one per prompt, and no second run for any prompt, so every consistency cell is empty.

That is exactly the kind of figure a tracker dashboard would happily chart as a trend. The definition of the metric the Index ranks by explains why a denominator of one is a placeholder and not a share.

Which tool should you pick at your size?

One product line, one category. Hand-run the checker template monthly, or subscribe to Otterly’s $29 Lite tier for 15 prompts on four engines if you want a chart. If the Index covers your category, read it and spend the money on the pages it says carry your mentions; total outlay is under $50 a month either way.

Multi-product vendor, three to six categories. The prompt count is what matters now, and six categories at 25 prompts is 150 prompts. That is Otterly Premium territory at $489 a month, or Profound’s $399 tier plus prompts beyond its 100, for which the page publishes no price.

This is also the size at which the DIY route starts to pay. At three runs, 150 prompts across four engines is 1,800 requests a month, under $40 in fees once a $25 SerpApi plan is included. The extraction code you wrote for one category runs on six.

Pick DIY if you have a developer who owns it; pick Otterly if you do not.

Agency. You need multi-brand, per-client exports and a login the client can hold. Profound’s Enterprise tier lists multiple companies tracked, Peec has an agencies section, and Otterly’s Standard plan lists unlimited team members.

The question to ask each is whether the client owns the run history when they leave you, which a DIY pipeline solves by design.

Limits

This comparison rests on what three vendors wrote on their own pages on one day. It has no hands-on trial of any tool, no run-level data from any of them, and no way to verify a claimed engine count or cadence. Peec’s absence of published prices is a fact about its pricing page, not about its product.

Prompt counts, add-on prices and tier names change; a fetch six months from now will disagree with this table somewhere. The DIY figures are arithmetic on published API prices at the Index’s sample size and exclude engineering time entirely. A tracker with no one reading it is a cron job.

The Index itself is small, covers property-ops only, and has one run per prompt on each of two engines. Nothing here tells you which tool produces numbers closer to what a real buyer sees, because no tool, including the Index, sees that.

What changes month to month?

Vendor pricing and tier contents, which is why the research file behind this page carries a fetch date on every row. Engine lists and add-on structures, which differ by tier today and are the part of a pricing page most likely to change. API prices and free quotas, which move the DIY arithmetic.

The models behind each engine, which is why the only tool worth trusting on a trend is one that stores the model identifier per run. And the Index’s own coverage, which grows by engine, category and prompt set version, each logged on the Index changelog with the edition it takes effect.

Related guides

• AI visibility: the measurement framework, the hub above this page, which defines every Index metric.
• Index methodology, the source of truth for prompt design, capture, extraction and the metric table.
• The metric that replaces share of voice, for the definition and the arithmetic behind a mention rate.
• How AI Overviews are triggered and sourced, for why SERP capture is the hard part of any tracker.
• the ChatGPT sources guide, for the engine every tool covers first.
• Monthly monitoring, the practice offer that reports Index figures per edition.

Changelog

• 2026-10-28: Reformatted with a short answer, question headings, the evaluation checklist as a numbered list, and an FAQ. Retitled toward “best AI visibility tools”. The Index paragraph now says ChatGPT is the only measured engine. Vendor facts unchanged from the 2026-09-04 fetch.
• 2026-09-26: The September figures now cite the 25 ChatGPT app captures per category; the API run of 3 September 2026 is described as retired from the edition and archived.
• 2026-09-19: Corrected the consumer-product gap section. The Index now captures ChatGPT from the app rather than the API, with memory and history off. The 2026-09 run is labeled as the API capture of 3 September 2026.
• 2026-09-04: First published.