Key takeaways
- Tracking tells you where you stand. It does not fix anything. The 30 days after your first baseline is where rank actually moves.
- ChatGPT cites product pages (22.4%) and landing pages (16.8%) more than any other content type, so start with the pages that describe what you sell.
- Almost half of ChatGPT responses cite 10 or more domains, so you win by being in the mix, mostly through third-party lists, reviews and comparisons.
- Check crawler access first. A beautiful page that OAI-SearchBot cannot read will never be cited.
- Re-measure across many prompts and repeated runs. One check is one sample, not a result.
Why tracking alone does not move your ChatGPT rank
You installed a tracker, ran a prompt set, and got a dashboard with a low visibility number and a few competitors sitting above you. That is a good start. It is also where most teams stall, because a dashboard is a diagnosis and nobody has written the prescription.
Here is the plan I would run over the next 30 days. It is built on citation data rather than on folklore about prompt tricks. The short version: figure out which pages ChatGPT cites in your category, make sure your own pages can be read and are complete, get named on the third-party pages that already get cited, and then measure again.
One caveat before you start. ChatGPT does not return ten blue links, so "rank" really means three things: whether you are mentioned, whether your domain is cited, and where you land among the other brands in the answer. Searchable reports that one ChatGPT response is backed by an average of 1.74 underlying searches, across 6.9 million responses, which means any single check is one sample. Omnia makes a similar point: the same prompt can vary by run, model version and locale. So everything below assumes you look at trends across a prompt set, not at one lucky or unlucky answer.
What the citation data says you should fix first
Before picking tasks, look at what ChatGPT actually cites. Promptwatch Data analyzed ChatGPT citations from Sep 28 to Oct 4, 2026 and found this split:
| Content type | Share of ChatGPT citations |
|---|---|
| Product pages | 22.4% |
| Landing pages | 16.8% |
| Listicles | 13.6% |
| How-tos | 8.7% |
| Docs | 7.5% |
| General articles | 6.1% |
| Comparisons | 5.9% |
| News | 5.7% |
| FAQ | 3.3% |
| Reviews | 1.7% |
Two things stand out. First, product and landing pages together take about 39% of citations. If your product page is thin, out of date, or a wall of marketing adjectives, that is the most expensive gap on your site. Second, the intent matters. For commercial prompts, product pages hit 26.0% and landing pages 20.6%. For branded prompts, reviews jump to 5.6%, more than three times their overall share.
Then there is the crowding problem. In the Promptwatch report on unique domains per response, 47.6% of cited responses cite 10 or more domains, and 66.1% cite at least 8. Only 6% cite three or fewer. You are never alone in an answer. For competitor comparison prompts, 49.1% cite 10 or more domains. For brand-specific prompts it drops: only 21.3% cite 10 or more, and the most common count is 5.
The practical read: for category prompts you need to be one of many sources, which usually means getting into other people's listicles and reviews. For brand prompts, a handful of review sites and forums shape how ChatGPT describes you, so those few pages matter a lot.
Before day 1: set up a baseline you can trust
Skip this and you will not know in week four whether anything worked.
Pick 30 to 50 prompts that your buyers actually type. Split them into three groups that mirror the data above: commercial ("best X for Y"), informational ("how do I do Z"), and branded ("is [your brand] good for..."). Add a handful of competitor comparison prompts. Ask ChatGPT itself to suggest variants, then keep the ones that sound like real people.
Run the set for at least a week before you change anything. Record four numbers per group: how often you are mentioned, how often your own domain is cited, your average position among named brands, and which competitors appear next to you. If your tracker supports it, also capture the sub-queries ChatGPT runs behind each answer, because those query fan-outs tell you exactly which searches your pages need to match. Promptwatch's query fan-out report explains how many searches ChatGPT triggers per response.
Picking a tracker
You have plenty of options. The useful distinction is between tools that report and tools that help you act.
| Tool | Strength | Watch out for |
|---|---|---|
| Promptwatch | Tracking plus crawler logs, citation analysis, content gaps, a Content Agent and an action list | More than you need if you only want a score |
| Searchable | Covers multiple ChatGPT surfaces with content and technical tools | Check plan limits against your prompt count |
| Otterly.AI | Low-cost monitoring | Basic; little help with the fixes |
| Peec AI | Clean visibility tracking | Monitoring focus |
| Rankscale | Frequent sampling | Narrower on execution |
| Morningscore | Beginner friendly | Lighter on citation depth |
Promptwatch is the one I would pick for this plan specifically, because the plan needs more than a number. It monitors the real ChatGPT interface, shows which of your pages get cited and where they land in the source list, logs AI crawler hits on your site, and turns gaps into a prioritized to-do list. You can start with the free AI visibility report or the 7-day trial on a paid plan; confirm current prices at promptwatch.com/pricing.
If you want to compare more options, the GEO software directory lists dozens of them.
Week 1: fix access and the pages that should already be winning
Start with the boring stuff. It is the cheapest, and it is where hidden losses sit.
Check that ChatGPT's crawlers can reach you
Open your robots.txt and look for rules that block OAI-SearchBot, ChatGPT-User or GPTBot. Blocking GPTBot only stops training use; blocking OAI-SearchBot means ChatGPT Search cannot index you. Many sites block all three by accident through a blanket bot rule or a CDN firewall setting. Also check that important pages return a 200, render without needing a login, and do not hide their core text inside scripts that bots skip.
If you can see server logs, look at which AI bots actually hit your pages and which return errors. Promptwatch's AI crawler logs do this with IP-verified requests, so spoofed bots are filtered out. Free options: the AI Crawler URL Inspector and the Robots.txt Generator for AI. A page that is cited but crawled rarely is an access signal. A page that is crawled often but never cited is a content problem. Knowing which one you have saves weeks.
Rewrite your product and landing pages
Take your top five product and landing pages and make each one complete and factual. What that means in practice:
- State what the product is, who it is for, and what it costs, in plain sentences near the top.
- Put specs, plans, integrations and limits in text, not only in images or accordions.
- Add the comparison facts buyers ask about: how it differs from the main two or three alternatives.
- Use the same language your prompts use. If buyers type "CRM for small agencies", a page that only says "relationship platform" misses the match.
- Date-stamp pricing and update it when it changes.
The half-life data gives a reason to be careful here. In the Promptwatch content half-life report, landing pages and docs held the longest: 78% of landing pages and 76% of docs cited on Jul 6 were still cited 30 days later, and about two-thirds were still cited after 60 days. Product pages, reviews and how-tos sat closer to 55% after 60 days, with half-lives of 74 to 77 days. Landing pages and docs are long-term assets. Product pages need regular refreshing.
Week 2: fill the content gaps that match real prompts
By now your baseline has a clear shape. Find the prompts where you have no mention and no citation, and ask what kind of page ChatGPT is citing in your place.
Map each gap to an intent:
- Commercial gap: you need a product or landing page, a comparison page, or a spot in a listicle.
- Informational gap: you need a how-to, a doc page or an FAQ.
- Branded gap: you need reviews and third-party profiles that say accurate things.
Then build the pages in order of value. Comparison pages are underrated: they only make 5.9% of citations overall, but they showed 72% of pages still cited after 30 days, and competitor comparison prompts almost never cite a single domain (0.35%), so there is room to be one of many. Publish your own honest comparison against the two or three alternatives buyers mention. Add a public, crawlable FAQ and docs where buyers ask practical questions.
A content gap analysis does this faster than guessing. Promptwatch maps your pages against the prompts AI answers and scores coverage, and its Content Agent can draft and publish to WordPress, Webflow or Framer from those gaps; see the Content Agents page. If you would rather keep writing in-house, a content optimizer can still help you structure pages against the cited competitors.
One rule for AI-assisted drafts: edit them. ChatGPT cites pages that state facts clearly. Generic filler written to hit a keyword list is exactly what gets skipped.
Week 3: get named on the pages ChatGPT already trusts
Owned pages are half the job. Because most answers cite many domains, third-party pages are often how a brand gets into the answer at all.
Find the pages that are cited for your prompts
List the top cited domains and URLs across your prompt set. You will usually see the same handful of roundups, review sites and comparison pages. Each one is a target. Check whether you are listed, whether the description is right, and whether a competitor has a better entry.
Industry matters here. In Promptwatch's software citation report (Sep 28 to Oct 4, 2026), GitHub, Microsoft's docs and G2 lead, and documentation and help centers earn the largest share. In the travel report, booking and search sites earn more than half. In automotive, marketplaces like Autotrader and Cars.com outrank every car brand. Your category has its own pattern, so go to the report closest to yours before deciding where to spend effort.
Do the outreach
For each target page, pick one move: ask to be added to a listicle, supply facts and screenshots for a review, collect reviews on the sites buyers trust, or pitch an update to an outdated comparison. Digital PR helps for news-style visibility, but remember that news fades fast: half of cited news pages were gone after 36 days in the half-life data. Use press for short bursts, not as the foundation.
A note on Reddit. Promptwatch's data shows reddit.com's share of ChatGPT citations fell from about 4% to 0.5% on August 14, 2026, and ChatGPT drew only 0.46% of its citations from social platforms in the Sep 28 to Oct 4 window, against 12.2% for Google AI Overviews. If your plan leans on Reddit threads for ChatGPT, rethink it. Reddit still matters for other AI engines and for human buyers, just not as a ChatGPT lever.
If you want a human team on outreach and AI search work, an agency like 1001 SEO Media covers digital PR, content and GEO together. That is optional; the plan works in-house too.
Week 4: fix the story, then re-measure
Check what ChatGPT says about you
Read the actual claims in your branded answers. Is the pricing right? Does it say you lack a feature you shipped last year? Find the page ChatGPT learned it from, and fix that page. Often it is an old review, an outdated listing, or a stale page on your own site. Promptwatch's Brand Narrative pulls these claims out of every response, and Fact Check on Business plans traces inaccurate ones back to the pages that state them.
Re-run and compare
Run the same prompt set you used for the baseline, at the same cadence, in the same country and language. Compare the four numbers: mention rate, own-domain citation rate, average position, and which competitors share your answers. Look at trends by prompt group, not individual prompts. Expect movement in owned-page citations first, since access fixes and page rewrites are within your control. Third-party placements take longer.
Also check that traffic follows. Tag AI-referred visits in analytics. ChatGPT stopped sending a referrer on October 8, 2026, so use a utm_source=chatgpt.com setting where your analytics tool supports it, or an AI-aware analytics layer. A rise in citations with no change in visits means you are cited but not clicked, which is a page-snippet problem, not a visibility one.
The 30-day checklist
| Week | Task | Done when |
|---|---|---|
| Before | Build prompt set, run 7-day baseline | You have mention, citation and position numbers by intent |
| 1 | Audit robots.txt, CDN rules and crawl errors | OAI-SearchBot reads your key pages with no errors |
| 1 | Rewrite top 5 product and landing pages | Specs, prices and positioning are complete and dated |
| 2 | Map gaps to intent, publish comparison, FAQ and docs | Every priority gap has a page assigned or live |
| 3 | Pitch the top cited roundups and review sites | At least 5 placements requested or secured |
| 4 | Correct inaccurate claims at the source | Wrong facts removed from your own pages and listings |
| 4 | Re-run prompt set and compare | You have a before/after on all four numbers |
What to expect, and what to do next
Be realistic. Thirty days is enough to fix access, rebuild your best pages, and start placements. It is not enough to dominate a competitive category. I would treat the first month as the cleanup, and then keep a weekly rhythm: check the trend, pick two actions, ship them, and refresh product pages every couple of months since they decay faster than landing pages.
If the number barely moves after a month, check in this order: crawler access, page completeness, then third-party coverage. It is almost always one of those three, and in that order of likelihood.
For more background on how ChatGPT chooses sources, browse the Promptwatch Data reports, or the Promptwatch Academy for metric definitions and a 90-day GEO program that extends this plan.