Troubleshooting
Six things that trip people up, with the real reason behind each one and exactly what to do about it. Most of these aren't bugs — they're the product behaving correctly in a way that isn't obvious the first time you see it.
A prompt run failed, or looks stuck
A tracking batch is every active prompt asked of every tracked platform, run as a set of individual jobs. Here’s what a few common symptoms actually mean:
| Symptom | Likely cause | What to do |
|---|---|---|
| “Run tracking now” is disabled or says a run is already in progress | Only one batch runs per project at a time — a guard against double-starting. | Wait for the current batch to finish; the header shows a live progress bar while it runs. |
| The progress bar hasn’t moved in a long time | A batch still “pending” or “running” after 30 minutes is treated as orphaned by a crashed process, not a live run. | Nothing to do — the next time you load the page (or try to start a new run), OptimizeCamp reports it as failed automatically and the button comes back. Start a fresh run. |
| The batch finished, but shows “Partial” instead of “Completed” | One or more individual prompt×platform jobs hit a transient error while the rest of the batch completed normally. | Check the Prompts table for rows still showing “No answer” on their latest run, then start a new tracking run — every prompt is re-asked from scratch each batch, so a fresh run retries anything that failed before. |
| A prompt shows “No answer” on ChatGPT specifically | An actual failure — unlike Google AI Overviews (see below), a successful ChatGPT run essentially always returns text. | Re-run tracking. If it keeps happening only on this one prompt across several runs, it’s worth flagging to your team. |
Note
If the dashboard tells you there are no active prompts to run, that’s not a failure — every prompt in the project is paused or archived. Reactivate at least one from Managing tracked prompts and try again.
No Google AI Overview appeared for a prompt
You’ll sometimes see a prompt with a real ChatGPT answer but “No answer” in the Google AI Overviews column for that same run.
This is expected, and it’s meaningful, not broken. Google doesn’t generate an AI Overview box for every search — for a real share of queries it shows classic results instead. When that happens, OptimizeCamp records it honestly as “no answer” rather than forcing a result that didn’t exist. That absence is itself useful data: it tells you this particular question doesn’t currently surface an AI Overview at all, so there’s nothing to be visible or invisible within on that platform, for that run.
Tip
How to tell this apart from a real failure: a legitimate no-trigger only ever shows up on the Google AI Overviews column. If ChatGPT is also repeatedly showing “No answer” on the same prompt, that’s the actual problem worth investigating — see above.
There’s nothing to fix on a legitimate non-trigger — it’s simply re-checked on your next scheduled or on-demand tracking run, and some prompts may show an Overview on one run and not the next. See Answer engines we track.
WordPress connection error or a publish conflict
Connection errors
OptimizeCamp writes back specific, readable error messages when a WordPress connection or sync fails — here’s what the most common ones mean:
| Message you see | What it means | What to do |
|---|---|---|
| “WordPress rejected the connection” | Wrong username or application password, or the password was revoked or regenerated on the WordPress side. | Create a fresh application password in wp-admin under Users › Profile › Application Passwords, then reconnect from Settings and paste it exactly. |
| “REST API unreachable” | /wp-json is blocked — a security plugin (Wordfence, iThemes), a firewall rule, or permalinks set to “Plain” with no rewrite rules. | Check the plugin’s REST API allowlist, or visit yoursite.com/wp-json/wp/v2/ directly in a browser — if that 404s or shows HTML, the API itself is blocked, not just OptimizeCamp’s connection to it. |
| “WordPress redirected the request” | The saved site address doesn’t match the site’s real one — http vs. https, or www vs. no-www. | Reconnect using the exact address the site actually resolves to. |
| “This user can’t edit posts” / “can’t publish posts” | The connected account’s role is missing a needed capability. | Reconnect with an Editor or Administrator account. |
| Couldn’t reach the site / timed out | The site is down, overloaded, or behind a firewall blocking outside requests. | Confirm the site is reachable from outside your network, then try Sync now again. |
Full connection setup, including exactly what an application password is and which role to use, is covered in Connecting WordPress.
Publish conflicts
A page shows status Conflict instead of Synced or Published. This means the post changed on WordPress and inside the OptimizeCamp editor since the last sync. Rather than guessing which version should win, OptimizeCamp blocks the publish, leaves your local draft untouched, and waits for you to decide.
What to do
Re-sync from Settings › Sync now to pull the current WordPress version, then decide: re-apply any accepted fixes you still want against the fresh copy, or discard your local edit if the WordPress-side change should win instead.
Two related messages you might see: “This post no longer exists on WordPress” means it was deleted on the WordPress side — a sync will archive it in your inventory automatically, so run one to reconcile your library. And “Generated drafts can’t update WordPress directly” means you’re trying to push an update to a generated article that’s still sitting as a WordPress draft — publish it from wp-admin first (mandatory human review for generated content), and updates work normally after that.
A competitor or your brand isn’t being detected
You read an answer yourself and can plainly see your brand (or a competitor) named, but the Prompts table or dashboard shows Absent, or a competitor’s numbers look understated.
Matching is exact and word-boundary, not fuzzy or semantic — it checks the answer text for any of the aliases on file for that brand or competitor, case-insensitively, but only an exact spelling. If the AI used a variant — an abbreviation, a product name instead of the company name, punctuation stuck to the word — and that exact variant isn’t in your alias list, it’s invisible to the matcher.
There’s a second, subtler cause on real (non-simulated) answers: a small language model double-checks every text match to filter out a same-spelled word that clearly isn’t your brand (a dictionary word, a generic term). That disambiguation step can occasionally misjudge a genuine mention, though it defaults to keeping the mention whenever the check itself fails.
What to do
Open your brand profile (or the competitor’s row on Competitors) and add every real-world variant you can think of — misspellings, abbreviations, “&” vs. “and,” or a product name your brand goes by. Getting aliases right is the single highest-leverage fix for missed detection.
Separately: a brand that keeps showing up in your answers but was never added as a tracked competitor won’t appear in Share of Voice or anywhere else. Check the “untracked brands appearing in your answers” list on Competitors and promote it with one click if it’s a real rival.
A number looks off, or jumped suddenly
Visibility Score, Mention Rate, or Share of Voice moved more than you expected between visits, or doesn’t match what you remember. Four real causes, roughly in order of likelihood:
- You’re looking at fresh data. Every number is computed live from stored answers each time the page loads — there’s no separate “refresh” step. If a tracking run finished since you last looked, what you see now already includes it.
- Small sample size. With a modest number of tracked prompts, one answer flipping from mentioned to absent (or back) moves a percentage-based metric like Mention Rate by several points on its own — the fewer runs behind a number, the more a single answer can swing it.
- An alias was added or removed. Editing your brand profile changes what counts as a match going forward. A jump right around that edit likely reflects the new matching rules catching (or missing) mentions the old rules didn’t — not a sudden real-world shift in how AI talks about you.
- Google AI Overview trigger rate shifted. A run where Google doesn’t return an Overview simply isn’t counted in that platform’s scope for the day — see above. A batch with fewer triggers can shift the combined-platform number even if ChatGPT looked identical.
What to do
Click into the specific prompt behind the swing and read the actual answer text for the runs on either side of the change — that’s always the fastest way to confirm whether a number reflects something real. See Reading an answer.
You’re out of audit or article credits
The Audit this page button, or Generate article, is locked or unavailable.
Audits and article generations are metered per plan — one audit credit per audit run (refunded automatically if the audit fails or stalls partway — a completed audit consumes the credit even in keyless fallback), one article credit per full generation (also refunded on failure; regenerating a single section for free doesn’t count). Scheduled tracking runs are never credited — those are governed by your plan’s cadence and prompt cap instead, a separate limit entirely.
Note
Running out doesn’t stop you from working — the Priority Score, QuickScore, and the Optimization Queue itself are always free and keep recomputing, so you can keep triaging what to fix next even at zero credits. You just can’t start a new deep audit or generate a new draft until your next cycle or an upgrade.
See Plans, quotas & credits for exact allotments per tier and when they reset.
FAQ
Straight answers to the questions new users ask most often.
Glossary
Every term in one place, defined in plain language, from AEO to Visibility Score.
Plans, quotas & credits
How prompts, audit credits and article credits work — and what happens when you hit a limit.
Your brand profile
Brand name, aliases, topics and competitors — the profile that powers accurate measurement.
Still stuck?
Most detection and measurement issues trace back to two places: your brand profile (aliases especially) and the actual answer text behind a number. When in doubt, go read the real answer — it’s always one click away.