Profound for AI Visibility: What the Dashboard Shows and What We Do With It

What Profound Is (and Isn’t)

Profound is an AI visibility platform. It runs prompts across the major models, captures how each one answers, records which sources get cited, and tracks how often a brand is mentioned versus its competitors. If you want to know whether AI talks about you — and how you stack up — Profound is the instrument that measures it.

What it isn’t is a diagnosis. Profound shows where a brand stands across AI. It doesn’t tell you why it’s there or what to do about it. That gap — between “visibility is 4%” and “here is the homepage sentence to rewrite, the three publications to pitch, and the Reddit thread feeding the wrong answer” — is where we live.

Every engagement starts with the full Profound dataset. Every prompt, every AI response, every citation, every mention across platforms. We pull all of it. The output isn’t a report that says “visibility is low.” It’s a scored breakdown of 48 things that matter, ranked by impact, with specific fixes attached.

How We Think About AI Visibility

Most agencies treat AI visibility like SEO with a new name. Write more, add schema, hope it works. That misses the point. AI doesn’t rank pages. It pulls sentences, checks whether independent sources agree, and decides whether to recommend you outright or hedge with “according to their website…”

The difference between “Brand X is the leading provider” and “Brand X claims to be a leading provider” isn’t word choice. It’s whether AI found anyone else who said the same thing. That’s a structural problem. You can’t blog your way out of it.

So we don’t do content marketing with an AI label on it. We fix how AI systems understand a brand before we scale anything. That matters because if AI doesn’t know who you are, more content just gives it more material to get wrong.

The Six Layers We Score

We score brands across six layers. Each one is a different way AI can get you wrong, and each one needs a different fix.

1. Access — Can AI crawlers reach your content?

If robots.txt blocks GPTBot or the site returns 403s to ClaudeBot, nothing else matters. We test every client against six bot user agents before doing anything else.

2. Identity — Does AI know who you are?

Not your name — does it understand what you do, who you serve, what category you belong in? We measure this across every major model. Gemini might get it right while Copilot thinks you’re a different company entirely. One client — a B2B commercial furniture manufacturer — was being described by AI as “an Amazon furniture retailer.” That’s not a small miss. That sends every recommendation sideways.

3. Language — Does AI use your words or make up its own?

We track echo rates and identify where your messaging sticks versus where AI substitutes something generic. We also look at whether your site makes claims AI can actually cite. “We strive to deliver value” gives AI nothing. “We process 10,000 orders daily across 14 distribution centres” gives it something to repeat.

4. Corroboration — Does anyone else say what you say?

The layer most agencies skip. AI doesn’t decide if your claims are true. It checks whether anyone else said the same thing. Your website says “we’re the leading provider” and nobody else does? AI hedges. Three editorial sources and a Reddit thread agree? AI states it as fact. We map which claims have backup and which are unsupported, then build the corroboration that’s missing.

5. Structure — Can AI find your point?

AI doesn’t read pages like people do. It pulls sentences. Google’s grounding pipeline works in roughly 500-token chunks and covers about a third of an average page. If your main point is in paragraph six of a 3,000-word article, AI won’t find it. We restructure content so the important parts are where AI looks first — answer-first formatting, standalone sentences, proper schema, internal links that connect topics the way AI expects.

6. Integrity — Does AI trust you?

The trust layer. Does AI cite real sources when it mentions you? Does it contradict itself across platforms? Does it state things about you with confidence or add qualifiers? We measure hallucination rates, citation quality, and how often AI asserts versus hedges. “Brand X is known for…” versus “Brand X claims to be…” — that gap tells you whether the trust signals are working.

What Every Engagement Produces

Every engagement produces a scored audit across all 48 metrics, a prioritised action plan, and a deliverable package:

• Executive summary with visibility broken down by platform, topic, and competitive position.
• Language recommendations showing where the brand’s messaging diverges from how AI describes it — current versus target, side by side.
• 12–15 ready-to-use text blocks with character counts and placement guidance, ready to drop into the site.
• An outreach brief with the specific publications, listicles, and directories where the brand needs to appear — ranked by how often AI cites them and whether competitors are already there.
• A follow-up & readiness report on what questions AI asks about the brand, what warnings it flags, and where the site is ready versus where it has gaps.
• A detailed scorecard walking through every metric with per-platform examples.

Every recommendation traces back to something specific in the data. Every action has a before and after. Your leadership can answer “how are we doing in AI?” with numbers.

Competitive Intelligence From Profound Data

SearchTides pulls the full dataset and builds competitive co-mention matrices. Not just “how often does the brand appear” but who else shows up in the same response, how often, and who wins when AI picks one.

For one furniture client, 76% of AI responses mentioning the brand also mentioned a specific competitor. That’s not a visibility problem. That’s AI treating them as interchangeable. More blog posts won’t fix it. Changing what makes the brand different, in the places AI actually checks, will.

Head-to-head analysis uses industry-specific language. Not generic “quality” and “value” terms — the words AI actually uses in each category, because AI describes car rental companies differently than it describes lending platforms.

We calculate win rates per descriptor. If a competitor wins on “affordable” but the client wins on “durable” and “commercial-grade,” that shapes positioning. The data shows where to compete and where to differentiate — based on what AI is saying right now, not what we think it should say.

Citation Source Mapping

Profound captures which URLs AI cites when mentioning a brand. We map every citation to its source domain and classify it: superfeeder (Reddit, YouTube, Wikipedia, LinkedIn, Forbes), editorial, directory, review platform, or brand-owned.

This shows which third-party sources are driving AI confidence in the brand — and which sources cite competitors but not the client. Those gaps become outreach targets with pitch angles pulled from the data. When AI cites a competitor on a listicle 200+ times and the client isn’t on that page, that’s an outreach problem with a measurable return.

We also track citation depth per engine. Being “cited on ChatGPT” can mean a lead recommendation with a direct link, or a passing mention at position five in a list. Profound data shows the brand appears. We show whether that appearance matters.

Listicle and Outreach Intelligence

Every URL in the Profound data goes through a listicle detection pipeline. Pages cited 50+ times that mention multiple brands in the category get flagged as opportunities. We calculate per-model citation rates, check whether the client is already listed, identify which competitors are there, and score outreach difficulty.

For one client, this turned up 573 specific pages from 1,946 AI responses. Each one came with which engines cite it, which competitors are on it, and how hard it would be to get included. That turns outreach from “let’s pitch some publications” into a ranked list with numbers behind it.

We verify brand presence on flagged pages using automated crawling. A URL might look like an opportunity, but the brand might already be listed and the data hasn’t caught up. Or the page went 404 six months ago. Verification keeps the pipeline clean.

Per-Model Intelligence

ChatGPT, Claude, Google AI Mode, Google AI Overviews, Gemini, Perplexity, Copilot — they don’t behave the same way. A brand can be well-known on ChatGPT and invisible on Google AI Mode. One favours conversational content. The other wants structured editorial sources. A single visibility number hides all of that.

We break it apart. Per-model reports show mention rates, own-site citations, competitive positioning, and recommendation patterns on each platform separately. When we say “focus on Google AI Mode,” it’s because the data shows that’s where the gap is and the audience is.

If Google AI Mode is the gap and it pulls from editorial sources, the strategy targets editorial placements. If Copilot is the gap and it indexes LinkedIn, the strategy targets LinkedIn publishing. Profound data tells us where to aim. We don’t guess.

Language and Sentiment Analysis

Profound captures how AI describes a brand — the specific words and framing. We analyse this at scale to find language gaps: where the brand says one thing and AI says something else.

We track sentiment proximity. Not just whether negative terms appear in AI responses about the brand, but whether they appear near the brand mention or somewhere else in the response. “Expensive” might show up, but if it’s 500 characters from the brand name, it’s about the category, not the brand. Proximity analysis separates real problems from noise. It prevents false alarms and keeps the action plan focused.

We also measure assertion confidence — how often AI states something about the brand as fact versus hedges. “Brand X is the leading provider” versus “Brand X describes itself as a leading provider.” The difference is whether AI found enough independent agreement to commit. Tracking this over time shows whether corroboration work is paying off. As independent sources validate the brand’s claims, the hedge rate drops.

Follow-Up and Warning Analysis

AI doesn’t answer once and stop. Users ask follow-ups. “What’s their track record?” “Do they offer bulk pricing?” “What are the downsides?” And AI flags concerns on its own — “their fund size is relatively small” or “their selection is more limited than competitors.”

We use Profound data to identify which follow-up questions AI generates about the brand and which warnings it surfaces.

For a venture capital client, AI asked about portfolio performance 67% of the time yet the site had almost nothing on fund performance. That’s a gap at the worst possible moment. For a furniture client, AI warned about “limited selection” but the brand has 10,000+ SKUs. The warning existed because the site didn’t make catalog breadth easy to extract.

We audit the brand’s content against each question and warning. Does the site answer this? Does it get ahead of this warning with a better frame? Brands that cover most of AI’s follow-ups and reframe most of its warnings control the conversation. Brands with gaps let AI fill them — usually with hedging, competitor mentions, or whatever it pulled from a three-year-old Reddit thread.

Off-Page Work We Handle Directly

The data points to the work. We do the work. SearchTides handles off-page execution directly, not as a referral:

• Reddit participation on the threads AI actually pulls from.
• Superfeeder placement — getting the brand onto the sources AI pulls from most: Reddit, YouTube, Wikipedia, LinkedIn, Forbes.
• Listicle outreach against the ranked target list from the Profound data.
• Directory submissions where competitors already appear and the client doesn’t.
• Editorial pitching with angles that come from the data, not from brainstorming.

Reporting and Measurement

We build reporting on top of Profound data so leadership can answer questions about AI performance without guessing. Weekly and monthly reports track visibility trends, competitive movement, citation changes, and the impact of specific actions.

Campaign impact analysis ties interventions — a listicle placement, a schema update, a homepage language change — to shifts in Profound metrics. When a client asks “did that Reddit strategy work?” the answer comes from the data.

Results

A deposit-rates marketplace went from 1.5% AI visibility to 8% — now mentioned alongside Bankrate, NerdWallet, and Investopedia, and winning. And 66% of our clients get promoted internally after working with us.

Results usually show within 30 to 60 days. Sometimes faster. The range depends on how much structural work is needed and how fast changes can go live. CreditNinja is a six-year partnership; our own brand took 28 days.

Who We Work With

Growing brands in financial services, SaaS, marketplaces, furniture, and ecommerce. Companies where traffic is flat, rankings look fine but revenue doesn’t match, or competitors keep showing up in AI responses that used to be theirs.

We specialise in regulated industries — financial services, healthcare, identity verification — where a hallucinated claim or a misattributed stat has real consequences beyond marketing.

Not the right fit for local businesses, very early startups, or brands that don’t have enough presence for AI to have formed an opinion yet. We need something to work with.

Why Us

We start with comprehension, not promotion. Fix how AI understands the brand, then scale how it gets recommended. Every recommendation traces back to something specific in the data. Every action has a before and after. Profound shows where a brand stands across AI. SearchTides figures out why it’s there and what to do about it.

FAQs