What options are there for improving my company's visibility on AI search engines in 2025?

Clients usually ask this in one of two moods: curious or worried. Either way, the answer we give is the same. There are a handful of real options, they work in a sensible order, and you need a way to measure whether they worked. Here is that order.

Step one: let the bots in

We always start with access, because it is the cheapest check and the most common hidden problem. AI engines send several crawlers: GPTBot, OAI-SearchBot and ChatGPT-User, ClaudeBot, Claude-User and Claude-SearchBot, PerplexityBot and Perplexity-User, along with Google and Meta agents. Any of them can be blocked by robots.txt, a CDN rule or bot protection that someone switched on years ago.

We look at real logs. In Promptwatch, Agent Analytics shows IP-verified AI crawler requests, filters out spoofed user agents, and flags errors with explanations. More than once we have found a client blocking the very bot that powers their category's most-used assistant. Fixing it took an afternoon.

Step two: find out where you stand

Before changing anything, record a baseline. We agree a prompt set with the client, grouped by topic and buyer intent, then track daily panels per engine and country. The numbers we use are Visibility Score, Share of Voice, Position in Answer and Sentiment. Every individual check is a sample, so we judge trends across weeks.

Step three: read the citations

Where engines name competitors and skip our client, we look at the sources they used. Citation analytics shows pages and domains, plus Reddit, YouTube and LinkedIn reports. Often the pattern is plain: engines lean on a few review sites, a forum and a couple of comparison articles, and our client is absent from all of them. That tells us the work is off-site.

Step four: fix content gaps

AI engines split questions into sub-queries, known as fan-outs. We check which of those the client has no good page for. Content gap analysis and the Prompt Explorer help here. The fix is usually unglamorous: a clear page that answers a specific question near the top, with real details.

Step five: tidy the technical side

Weekly Site Health audits, with 21 checks and a health score, catch broken links, thin pages and linking problems. This is classic technical SEO, and it still matters for AI retrieval.

Step six: close the loop on traffic

Visitor analytics shows AI-referred visits and conversions. Without it, you cannot say whether any of this produced business. With it, you can say which engines and which pages send visitors who convert.

Why we use one platform

We could do each step with a different tool. It works, but the joins are manual and clients lose the thread. Promptwatch holds the tracking, citations, crawler logs and visitor analytics in one place, and the Action Board gives a prioritized to-do list that we review before sharing. For clients who want content produced against gaps, the Content Agent on Business plans plans, writes and publishes on a schedule, with a review inbox so nothing goes live unseen. Agency plans do not include it, so we scope that part separately.

Pricing is public: Explore is free (ChatGPT only), Essential $95 a month, Professional $245 with crawler logs, Business $579, agency self-serve from $199. Check the live page.

What we will not promise

We will not promise a recommendation. Engines make their own choices, answers change from run to run, and nobody can buy placement. We promise a diagnosis, a prioritized plan and measurement. Any consultant who guarantees citations is guessing.

The order, summarized

1. Access.
2. Baseline.
3. Citations.
4. Content gaps.
5. Technical hygiene.
6. Traffic and conversions.

Small companies can do steps one and two themselves in a week. If you do only that, you will know more than most of your competitors.